diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index 1584bc76a9dd8ddff9d05a8cb693bcbd2e09fcde..b6a299ba4009ba73afc9673fe008b97ca139c57b 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -1,10 +1,11 @@ +repos: - repo: https://github.com/PaddlePaddle/mirrors-yapf.git - sha: 0d79c0c469bab64f7229c9aca2b1186ef47f0e37 + rev: 0d79c0c469bab64f7229c9aca2b1186ef47f0e37 hooks: - id: yapf files: \.py$ - repo: https://github.com/pre-commit/pre-commit-hooks - sha: a11d9314b22d8f8c7556443875b731ef05965464 + rev: a11d9314b22d8f8c7556443875b731ef05965464 hooks: - id: check-merge-conflict - id: check-symlinks @@ -15,7 +16,7 @@ - id: trailing-whitespace files: \.md$ - repo: https://github.com/Lucas-C/pre-commit-hooks - sha: v1.0.1 + rev: v1.0.1 hooks: - id: forbid-crlf files: \.md$ diff --git a/README.md b/README.md index a80b5f53c19fb642795c3dfee16a40d4ed9d9df5..44885f554afdc7e00188fae2987e7fbbb4278fcc 120000 --- a/README.md +++ b/README.md @@ -1 +1 @@ -README_ch.md +README_ch.md \ No newline at end of file diff --git a/README_ch.md b/README_ch.md index 41c1d44fd25dd435086e7d17505999587fccbce4..9219857fd950c4d5a4c96ae28ad80d7c5e060cb1 100644 --- a/README_ch.md +++ b/README_ch.md @@ -7,7 +7,7 @@ 飞桨图像识别套件PaddleClas是飞桨为工业界和学术界所准备的一个图像识别任务的工具集，助力使用者训练出更好的视觉模型和应用落地。 **近期更新** - +- 2022.4.21 新增 CVPR2022 oral论文 [MixFormer](https://arxiv.org/pdf/2204.02557.pdf) 相关[代码](https://github.com/PaddlePaddle/PaddleClas/pull/1820/files)。 - 2022.1.27 全面升级文档；新增[PaddleServing C++ pipeline部署方式](./deploy/paddleserving)和[18M图像识别安卓部署Demo](./deploy/lite_shitu)。 - 2021.11.1 发布[PP-ShiTu技术报告](https://arxiv.org/pdf/2111.00775.pdf)，新增饮料识别demo - 2021.10.23 发布轻量级图像识别系统PP-ShiTu，CPU上0.2s即可完成在10w+库的图像识别。 @@ -35,10 +35,11 @@ Res2Net200_vd预训练模型Top-1精度高达85.1%。 ## 欢迎加入技术交流群 -* 您可以扫描下面的微信群二维码， 加入PaddleClas 微信交流群。获得更高效的问题答疑，与各行各业开发者充分交流，期待您的加入。 +* 您可以扫描下面的QQ/微信二维码（添加小助手微信并回复“C”），加入PaddleClas微信交流群，获得更高效的问题答疑，与各行各业开发者充分交流，期待您的加入。
- + +
## 快速体验 diff --git a/README_en.md b/README_en.md index e86ed3fe1d8e39d8900b5f5da42711966295d61e..9b0d7c85d76cf06eac8fb265abb85c3bb98a275f 100644 --- a/README_en.md +++ b/README_en.md @@ -8,6 +8,8 @@ PaddleClas is an image recognition toolset for industry and academia, helping us **Recent updates** +- 2022.4.21 Added the related [code](https://github.com/PaddlePaddle/PaddleClas/pull/1820/files) of the CVPR2022 oral paper [MixFormer](https://arxiv.org/pdf/2204.02557.pdf). + - 2021.09.17 Add PP-LCNet series model developed by PaddleClas, these models show strong competitiveness on Intel CPUs. For the introduction of PP-LCNet, please refer to [paper](https://arxiv.org/pdf/2109.15099.pdf) or [PP-LCNet model introduction](docs/en/models/PP-LCNet_en.md). The metrics and pretrained model are available [here](docs/en/ImageNet_models_en.md). @@ -38,10 +40,11 @@ Four sample solutions are provided, including product recognition, vehicle recog ## Welcome to Join the Technical Exchange Group -* You can also scan the QR code below to join the PaddleClas WeChat group to get more efficient answers to your questions and to communicate with developers from all walks of life. We look forward to hearing from you. +* You can also scan the QR code below to join the PaddleClas QQ group and WeChat group (add and replay "C") to get more efficient answers to your questions and to communicate with developers from all walks of life. We look forward to hearing from you.
- + +
## Quick Start diff --git a/deploy/configs/inference_cartoon.yaml b/deploy/configs/inference_cartoon.yaml index 7d93d98cc0696d8e1508e02db2cc864d6f917d19..e79da55090130223466fd6b6a078b9909d6e26f2 100644 --- a/deploy/configs/inference_cartoon.yaml +++ b/deploy/configs/inference_cartoon.yaml @@ -8,7 +8,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 5 - labe_list: + label_list: - foreground use_gpu: True diff --git a/deploy/configs/inference_det.yaml b/deploy/configs/inference_det.yaml index c809a0257bc7c5b774f20fb3edb50a08e7d67bbb..dab7908ef7f59bfed077d9189811aedb650b0e92 100644 --- a/deploy/configs/inference_det.yaml +++ b/deploy/configs/inference_det.yaml @@ -5,7 +5,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 1 - labe_list: + label_list: - foreground # inference engine config diff --git a/deploy/configs/inference_drink.yaml b/deploy/configs/inference_drink.yaml index 61d116f9f407ca032e59132ec3971622c9df9a6e..1c3e2c29aa8ddd5db46bbc8660c9f45942696a9c 100644 --- a/deploy/configs/inference_drink.yaml +++ b/deploy/configs/inference_drink.yaml @@ -8,7 +8,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 5 - labe_list: + label_list: - foreground use_gpu: True @@ -51,8 +51,8 @@ RecPostProcess: null # indexing engine config IndexProcess: index_method: "HNSW32" # supported: HNSW32, IVF, Flat - index_dir: "./drink_dataset_v1.0/gallery" - image_root: "./drink_dataset_v1.0/index" + image_root: "./drink_dataset_v1.0/gallery" + index_dir: "./drink_dataset_v1.0/index" data_file: "./drink_dataset_v1.0/gallery/drink_label.txt" index_operation: "new" # suported: "append", "remove", "new" delimiter: " " diff --git a/deploy/configs/inference_general.yaml b/deploy/configs/inference_general.yaml index 6b397b5047b427d02014060380112b096e0b2da2..8fb8ae3a56697b882be00da554f33750ead42f70 100644 --- a/deploy/configs/inference_general.yaml +++ b/deploy/configs/inference_general.yaml @@ -8,7 +8,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 5 - labe_list: + label_list: - foreground use_gpu: True diff --git a/deploy/configs/inference_general_binary.yaml b/deploy/configs/inference_general_binary.yaml index d76dae8f8f7c70f27996f6b20fd623bdc00bc441..72ec31fc438d1f884bada59507a90d172ab4a416 100644 --- a/deploy/configs/inference_general_binary.yaml +++ b/deploy/configs/inference_general_binary.yaml @@ -8,7 +8,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 5 - labe_list: + label_list: - foreground use_gpu: True diff --git a/deploy/configs/inference_logo.yaml b/deploy/configs/inference_logo.yaml index f78ca25a042b3224a973d81f7b0242ace7c25430..2b8228eab772f8b1488275163518a6e059a49c53 100644 --- a/deploy/configs/inference_logo.yaml +++ b/deploy/configs/inference_logo.yaml @@ -8,7 +8,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 5 - labe_list: + label_list: - foreground use_gpu: True diff --git a/deploy/configs/inference_product.yaml b/deploy/configs/inference_product.yaml index e7b494c383aa5f42b4515446805b1357ba43107c..78ba32068cb696e897c39d516e66b323bd12ad61 100644 --- a/deploy/configs/inference_product.yaml +++ b/deploy/configs/inference_product.yaml @@ -8,7 +8,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 5 - labe_list: + label_list: - foreground # inference engine config diff --git a/deploy/configs/inference_vehicle.yaml b/deploy/configs/inference_vehicle.yaml index d99f42ad684150f1efeaf65f031ee1ea707fee37..e289e9f523b061dd26b8d687e594499dd7cdec37 100644 --- a/deploy/configs/inference_vehicle.yaml +++ b/deploy/configs/inference_vehicle.yaml @@ -8,7 +8,7 @@ Global: image_shape: [3, 640, 640] threshold: 0.2 max_det_results: 5 - labe_list: + label_list: - foreground use_gpu: True diff --git a/deploy/cpp_shitu/include/object_detector.h b/deploy/cpp_shitu/include/object_detector.h index 5bfc56253b1845a50f3b6b093db314e97505cfef..6855a0dcc84c2711283fe8d23ba1d2afe376fb0e 100644 --- a/deploy/cpp_shitu/include/object_detector.h +++ b/deploy/cpp_shitu/include/object_detector.h @@ -33,106 +33,106 @@ using namespace paddle_infer; namespace Detection { // Object Detection Result - struct ObjectResult { - // Rectangle coordinates of detected object: left, right, top, down - std::vector rect; - // Class id of detected object - int class_id; - // Confidence of detected object - float confidence; - }; +struct ObjectResult { + // Rectangle coordinates of detected object: left, right, top, down + std::vector rect; + // Class id of detected object + int class_id; + // Confidence of detected object + float confidence; +}; // Generate visualization colormap for each class - std::vector GenerateColorMap(int num_class); +std::vector GenerateColorMap(int num_class); // Visualiztion Detection Result - cv::Mat VisualizeResult(const cv::Mat &img, - const std::vector &results, - const std::vector &lables, - const std::vector &colormap, const bool is_rbox); - - class ObjectDetector { - public: - explicit ObjectDetector(const YAML::Node &config_file) { - this->use_gpu_ = config_file["Global"]["use_gpu"].as(); - if (config_file["Global"]["gpu_id"].IsDefined()) - this->gpu_id_ = config_file["Global"]["gpu_id"].as(); - this->gpu_mem_ = config_file["Global"]["gpu_mem"].as(); - this->cpu_math_library_num_threads_ = - config_file["Global"]["cpu_num_threads"].as(); - this->use_mkldnn_ = config_file["Global"]["enable_mkldnn"].as(); - this->use_tensorrt_ = config_file["Global"]["use_tensorrt"].as(); - this->use_fp16_ = config_file["Global"]["use_fp16"].as(); - this->model_dir_ = - config_file["Global"]["det_inference_model_dir"].as(); - this->threshold_ = config_file["Global"]["threshold"].as(); - this->max_det_results_ = config_file["Global"]["max_det_results"].as(); - this->image_shape_ = - config_file["Global"]["image_shape"].as < std::vector < int >> (); - this->label_list_ = - config_file["Global"]["labe_list"].as < std::vector < std::string >> (); - this->ir_optim_ = config_file["Global"]["ir_optim"].as(); - this->batch_size_ = config_file["Global"]["batch_size"].as(); - - preprocessor_.Init(config_file["DetPreProcess"]["transform_ops"]); - LoadModel(model_dir_, batch_size_, run_mode); - } - - // Load Paddle inference model - void LoadModel(const std::string &model_dir, const int batch_size = 1, - const std::string &run_mode = "fluid"); - - // Run predictor - void Predict(const std::vector imgs, const int warmup = 0, - const int repeats = 1, - std::vector *result = nullptr, - std::vector *bbox_num = nullptr, - std::vector *times = nullptr); - - const std::vector &GetLabelList() const { - return this->label_list_; - } - - const float &GetThreshold() const { return this->threshold_; } - - private: - bool use_gpu_ = true; - int gpu_id_ = 0; - int gpu_mem_ = 800; - int cpu_math_library_num_threads_ = 6; - std::string run_mode = "fluid"; - bool use_mkldnn_ = false; - bool use_tensorrt_ = false; - bool batch_size_ = 1; - bool use_fp16_ = false; - std::string model_dir_; - float threshold_ = 0.5; - float max_det_results_ = 5; - std::vector image_shape_ = {3, 640, 640}; - std::vector label_list_; - bool ir_optim_ = true; - bool det_permute_ = true; - bool det_postprocess_ = true; - int min_subgraph_size_ = 30; - bool use_dynamic_shape_ = false; - int trt_min_shape_ = 1; - int trt_max_shape_ = 1280; - int trt_opt_shape_ = 640; - bool trt_calib_mode_ = false; - - // Preprocess image and copy data to input buffer - void Preprocess(const cv::Mat &image_mat); - - // Postprocess result - void Postprocess(const std::vector mats, - std::vector *result, std::vector bbox_num, - bool is_rbox); - - std::shared_ptr predictor_; - Preprocessor preprocessor_; - ImageBlob inputs_; - std::vector output_data_; - std::vector out_bbox_num_data_; - }; +cv::Mat VisualizeResult(const cv::Mat &img, + const std::vector &results, + const std::vector &lables, + const std::vector &colormap, const bool is_rbox); + +class ObjectDetector { +public: + explicit ObjectDetector(const YAML::Node &config_file) { + this->use_gpu_ = config_file["Global"]["use_gpu"].as(); + if (config_file["Global"]["gpu_id"].IsDefined()) + this->gpu_id_ = config_file["Global"]["gpu_id"].as(); + this->gpu_mem_ = config_file["Global"]["gpu_mem"].as(); + this->cpu_math_library_num_threads_ = + config_file["Global"]["cpu_num_threads"].as(); + this->use_mkldnn_ = config_file["Global"]["enable_mkldnn"].as(); + this->use_tensorrt_ = config_file["Global"]["use_tensorrt"].as(); + this->use_fp16_ = config_file["Global"]["use_fp16"].as(); + this->model_dir_ = + config_file["Global"]["det_inference_model_dir"].as(); + this->threshold_ = config_file["Global"]["threshold"].as(); + this->max_det_results_ = config_file["Global"]["max_det_results"].as(); + this->image_shape_ = + config_file["Global"]["image_shape"].as>(); + this->label_list_ = + config_file["Global"]["label_list"].as>(); + this->ir_optim_ = config_file["Global"]["ir_optim"].as(); + this->batch_size_ = config_file["Global"]["batch_size"].as(); + + preprocessor_.Init(config_file["DetPreProcess"]["transform_ops"]); + LoadModel(model_dir_, batch_size_, run_mode); + } + + // Load Paddle inference model + void LoadModel(const std::string &model_dir, const int batch_size = 1, + const std::string &run_mode = "fluid"); + + // Run predictor + void Predict(const std::vector imgs, const int warmup = 0, + const int repeats = 1, + std::vector *result = nullptr, + std::vector *bbox_num = nullptr, + std::vector *times = nullptr); + + const std::vector &GetLabelList() const { + return this->label_list_; + } + + const float &GetThreshold() const { return this->threshold_; } + +private: + bool use_gpu_ = true; + int gpu_id_ = 0; + int gpu_mem_ = 800; + int cpu_math_library_num_threads_ = 6; + std::string run_mode = "fluid"; + bool use_mkldnn_ = false; + bool use_tensorrt_ = false; + bool batch_size_ = 1; + bool use_fp16_ = false; + std::string model_dir_; + float threshold_ = 0.5; + float max_det_results_ = 5; + std::vector image_shape_ = {3, 640, 640}; + std::vector label_list_; + bool ir_optim_ = true; + bool det_permute_ = true; + bool det_postprocess_ = true; + int min_subgraph_size_ = 30; + bool use_dynamic_shape_ = false; + int trt_min_shape_ = 1; + int trt_max_shape_ = 1280; + int trt_opt_shape_ = 640; + bool trt_calib_mode_ = false; + + // Preprocess image and copy data to input buffer + void Preprocess(const cv::Mat &image_mat); + + // Postprocess result + void Postprocess(const std::vector mats, + std::vector *result, std::vector bbox_num, + bool is_rbox); + + std::shared_ptr predictor_; + Preprocessor preprocessor_; + ImageBlob inputs_; + std::vector output_data_; + std::vector out_bbox_num_data_; +}; } // namespace Detection diff --git a/deploy/lite/Makefile b/deploy/lite/Makefile index f18864aa3ea983f9106a12ec784a848161a24a77..df57c8bb2a3b029331fbc9a046bea57fb5cd8fcc 100644 --- a/deploy/lite/Makefile +++ b/deploy/lite/Makefile @@ -1,9 +1,9 @@ ARM_ABI = arm8 export ARM_ABI -include ../Makefile.def +LITE_ROOT=./inference_lite_lib.android.armv8 -LITE_ROOT=../../../ +include ${LITE_ROOT}/demo/cxx/Makefile.def THIRD_PARTY_DIR=${LITE_ROOT}/third_party @@ -29,7 +29,7 @@ OPENCV_LIBS = ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/libs/libopencv_im ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/libtbb.a \ ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PATH}/3rdparty/libs/libcpufeatures.a -OPENCV_INCLUDE = -I../../../third_party/${OPENCV_VERSION}/${ARM_PATH}/include +OPENCV_INCLUDE = -I${LITE_ROOT}/third_party/${OPENCV_VERSION}/${ARM_PATH}/include CXX_INCLUDES = $(INCLUDES) ${OPENCV_INCLUDE} -I$(LITE_ROOT)/cxx/include diff --git a/deploy/lite/config.txt b/deploy/lite/config.txt index 08cee3dba2e73e2394a03e7730441dcd8dd8edb6..462fe8314508c7a6ad03a03796e1633e42cf0d3c 100644 --- a/deploy/lite/config.txt +++ b/deploy/lite/config.txt @@ -1,6 +1,11 @@ -clas_model_file ./MobileNetV3_large_x1_0.nb -label_path ./imagenet1k_label_list.txt +clas_model_file /data/local/tmp/arm_cpu/MobileNetV3_large_x1_0.nb +label_path /data/local/tmp/arm_cpu/imagenet1k_label_list.txt resize_short_size 256 crop_size 224 visualize 0 +num_threads 1 +batch_size 1 +precision FP32 +runtime_device arm_cpu enable_benchmark 0 +tipc_benchmark 0 diff --git a/deploy/lite/image_classfication.cpp b/deploy/lite/image_classfication.cpp index 535f3c86390581dfbe6e75fc2b2ad0f70408a838..3c97da661cc2e956e949e424fc83e92986d2cbb0 100644 --- a/deploy/lite/image_classfication.cpp +++ b/deploy/lite/image_classfication.cpp @@ -21,6 +21,7 @@ #include #include #include +#include "AutoLog/auto_log/lite_autolog.h" using namespace paddle::lite_api; // NOLINT using namespace std; @@ -149,8 +150,10 @@ cv::Mat CenterCropImg(const cv::Mat &img, const int &crop_size) { std::vector RunClasModel(std::shared_ptr predictor, const cv::Mat &img, const std::map &config, - const std::vector &word_labels, double &cost_time) { + const std::vector &word_labels, double &cost_time, + std::vector *time_info) { // Read img + auto preprocess_start = std::chrono::steady_clock::now(); int resize_short_size = stoi(config.at("resize_short_size")); int crop_size = stoi(config.at("crop_size")); int visualize = stoi(config.at("visualize")); @@ -172,8 +175,8 @@ RunClasModel(std::shared_ptr predictor, const cv::Mat &img, std::vector scale = {1 / 0.229f, 1 / 0.224f, 1 / 0.225f}; const float *dimg = reinterpret_cast(img_fp.data); NeonMeanScale(dimg, data0, img_fp.rows * img_fp.cols, mean, scale); - - auto start = std::chrono::system_clock::now(); + auto preprocess_end = std::chrono::steady_clock::now(); + auto inference_start = std::chrono::system_clock::now(); // Run predictor predictor->Run(); @@ -181,9 +184,10 @@ RunClasModel(std::shared_ptr predictor, const cv::Mat &img, std::unique_ptr output_tensor( std::move(predictor->GetOutput(0))); auto *output_data = output_tensor->data(); - auto end = std::chrono::system_clock::now(); + auto inference_end = std::chrono::system_clock::now(); + auto postprocess_start = std::chrono::system_clock::now(); auto duration = - std::chrono::duration_cast(end - start); + std::chrono::duration_cast(inference_end - inference_start); cost_time = double(duration.count()) * std::chrono::microseconds::period::num / std::chrono::microseconds::period::den; @@ -196,6 +200,13 @@ RunClasModel(std::shared_ptr predictor, const cv::Mat &img, cv::Mat output_image; auto results = PostProcess(output_data, output_size, word_labels, output_image); + auto postprocess_end = std::chrono::system_clock::now(); + std::chrono::duration preprocess_diff = preprocess_end - preprocess_start; + time_info->push_back(double(preprocess_diff.count() * 1000)); + std::chrono::duration inference_diff = inference_end - inference_start; + time_info->push_back(double(inference_diff.count() * 1000)); + std::chrono::duration postprocess_diff = postprocess_end - postprocess_start; + time_info->push_back(double(postprocess_diff.count() * 1000)); if (visualize) { std::string output_image_path = "./clas_result.png"; @@ -309,6 +320,12 @@ int main(int argc, char **argv) { std::string clas_model_file = config.at("clas_model_file"); std::string label_path = config.at("label_path"); + std::string crop_size = config.at("crop_size"); + int num_threads = stoi(config.at("num_threads")); + int batch_size = stoi(config.at("batch_size")); + std::string precision = config.at("precision"); + std::string runtime_device = config.at("runtime_device"); + bool tipc_benchmark = bool(stoi(config.at("tipc_benchmark"))); // Load Labels std::vector word_labels = LoadLabels(label_path); @@ -319,8 +336,9 @@ int main(int argc, char **argv) { cv::cvtColor(srcimg, srcimg, cv::COLOR_BGR2RGB); double run_time = 0; + std::vector time_info; std::vector results = - RunClasModel(clas_predictor, srcimg, config, word_labels, run_time); + RunClasModel(clas_predictor, srcimg, config, word_labels, run_time, &time_info); std::cout << "===clas result for image: " << img_path << "===" << std::endl; for (int i = 0; i < results.size(); i++) { @@ -338,6 +356,19 @@ int main(int argc, char **argv) { } else { std::cout << "Current time cost: " << run_time << " s." << std::endl; } + if (tipc_benchmark) { + AutoLogger autolog(clas_model_file, + runtime_device, + num_threads, + batch_size, + crop_size, + precision, + time_info, + 1); + std::cout << "=======================TIPC Lite Information=======================" << std::endl; + autolog.report(); + } + } return 0; diff --git a/deploy/lite/readme.md b/deploy/lite/readme.md index 176e62ceac8292071f5811e3c6e55972a0e3516b..473984dd3315fbc547fd83a1df57c68953b2fbbb 100644 --- a/deploy/lite/readme.md +++ b/deploy/lite/readme.md @@ -25,8 +25,8 @@ Paddle Lite是飞桨轻量化推理引擎，为手机、IOT端提供高效推理 1. [建议]直接下载，预测库下载链接如下： |平台|预测库下载链接| |-|-| - |Android|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv7.gcc.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv8.gcc.c++_static.with_extra.with_cv.tar.gz)| - |iOS|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| + |Android|[arm7](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.android.armv7.clang.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv.tar.gz)| + |iOS|[arm7](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| **注**： 1. 如果是从 Paddle-Lite [官方文档](https://paddle-lite.readthedocs.io/zh/latest/quick_start/release_lib.html#android-toolchain-gcc)下载的预测库， @@ -44,11 +44,11 @@ git checkout develop **注意**：编译Paddle-Lite获得预测库时，需要打开`--with_cv=ON --with_extra=ON`两个选项，`--arch`表示`arm`版本，这里指定为armv8，更多编译命令介绍请参考[链接](https://paddle-lite.readthedocs.io/zh/latest/user_guides/Compile/Android.html#id2)。 -直接下载预测库并解压后，可以得到`inference_lite_lib.android.armv8/`文件夹，通过编译Paddle-Lite得到的预测库位于`Paddle-Lite/build.lite.android.armv8.gcc/inference_lite_lib.android.armv8/`文件夹下。 +直接下载预测库并解压后，可以得到`inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/`文件夹，通过编译Paddle-Lite得到的预测库位于`Paddle-Lite/build.lite.android.armv8.gcc/inference_lite_lib.android.armv8/`文件夹下。 预测库的文件目录如下： ``` -inference_lite_lib.android.armv8/ +inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/ |-- cxx C++ 预测库和头文件 | |-- include C++ 头文件 | | |-- paddle_api.h @@ -86,7 +86,7 @@ Python下安装 `paddlelite`，目前最高支持`Python3.7`。 **注意**：`paddlelite`whl包版本必须和预测库版本对应。 ```shell -pip install paddlelite==2.8 +pip install paddlelite==2.10 ``` 之后使用`paddle_lite_opt`工具可以进行inference模型的转换。`paddle_lite_opt`的部分参数如下 @@ -146,6 +146,24 @@ paddle_lite_opt --model_file=./MobileNetV3_large_x1_0_infer/inference.pdmodel -- **注意**：`--optimize_out` 参数为优化后模型的保存路径，无需加后缀`.nb`；`--model_file` 参数为模型结构信息文件的路径，`--param_file` 参数为模型权重信息文件的路径，请注意文件名。 + + +#### 2.1.4 执行编译，得到可执行文件clas_system + +```shell +# 克隆 Autolog 代码库，以便获取自动化日志 +cd PaddleClas_root_path +cd deploy/lite/ +git clone https://github.com/LDOUBLEV/AutoLog.git +``` + +```shell +# 编译 +make -j +``` + +执行 `make` 命令后，会在当前目录生成 `clas_system` 可执行文件，该文件用于 Lite 预测。 + ### 2.2 与手机联调 @@ -167,7 +185,7 @@ paddle_lite_opt --model_file=./MobileNetV3_large_x1_0_infer/inference.pdmodel -- win上安装需要去谷歌的安卓平台下载ADB软件包进行安装：[链接](https://developer.android.com/studio) -4. 手机连接电脑后，开启手机`USB调试`选项，选择`文件传输`模式，在电脑终端中输入： +3. 手机连接电脑后，开启手机`USB调试`选项，选择`文件传输`模式，在电脑终端中输入： ```shell adb devices @@ -178,40 +196,18 @@ List of devices attached 744be294 device ``` -5. 准备优化后的模型、预测库文件、测试图像和类别映射文件。 - -```shell -cd PaddleClas_root_path -cd deploy/lite/ -# 运行prepare.sh -# prepare.sh 会将预测库文件、测试图像和使用的字典文件放置在预测库中的demo/cxx/clas文件夹下 -sh prepare.sh /{lite prediction library path}/inference_lite_lib.android.armv8 +4. 将优化后的模型、预测库文件、测试图像和类别映射文件push到手机上。 -# 进入lite demo的工作目录 -cd /{lite prediction library path}/inference_lite_lib.android.armv8/ -cd demo/cxx/clas/ - -# 将C++预测动态库so文件复制到debug文件夹中 -cp ../../../cxx/lib/libpaddle_light_api_shared.so ./debug/ -``` - -`prepare.sh` 以 `PaddleClas/deploy/lite/imgs/tabby_cat.jpg` 作为测试图像，将测试图像复制到`demo/cxx/clas/debug/` 文件夹下。 -将 `paddle_lite_opt` 工具优化后的模型文件放置到 `/{lite prediction library path}/inference_lite_lib.android.armv8/demo/cxx/clas/debug/` 文件夹下。本例中，使用[2.1.3](#2.1.3)生成的 `MobileNetV3_large_x1_0.nb` 模型文件。 - -执行完成后，clas文件夹下将有如下文件格式： - -``` -demo/cxx/clas/ -|-- debug/ -| |--MobileNetV3_large_x1_0.nb 优化后的分类器模型文件 -| |--tabby_cat.jpg 待测试图像 -| |--imagenet1k_label_list.txt 类别映射文件 -| |--libpaddle_light_api_shared.so C++预测库文件 -| |--config.txt 分类预测超参数配置 -|-- config.txt 分类预测超参数配置 -|-- image_classfication.cpp 图像分类代码文件 -|-- Makefile 编译文件 +```shell +adb shell mkdir -p /data/local/tmp/arm_cpu/ +adb push clas_system /data/local/tmp/arm_cpu/ +adb shell chmod +x /data/local/tmp/arm_cpu//clas_system +adb push inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/cxx/lib/libpaddle_light_api_shared.so /data/local/tmp/arm_cpu/ +adb push MobileNetV3_large_x1_0.nb /data/local/tmp/arm_cpu/ +adb push config.txt /data/local/tmp/arm_cpu/ +adb push ../../ppcls/utils/imagenet1k_label_list.txt /data/local/tmp/arm_cpu/ +adb push imgs/tabby_cat.jpg /data/local/tmp/arm_cpu/ ``` #### 注意： @@ -224,32 +220,22 @@ clas_model_file ./MobileNetV3_large_x1_0.nb # 模型文件地址 label_path ./imagenet1k_label_list.txt # 类别映射文本文件 resize_short_size 256 # resize之后的短边边长 crop_size 224 # 裁剪后用于预测的边长 -visualize 0 # 是否进行可视化，如果选择的话，会在当前文件夹下生成名为clas_result.png的图像文件。 +visualize 0 # 是否进行可视化，如果选择的话，会在当前文件夹下生成名为clas_result.png的图像文件 +num_threads 1 # 线程数，默认是1。 +precision FP32 # 精度类型，可以选择 FP32 或者 INT8，默认是 FP32。 +runtime_device arm_cpu # 设备类型，默认是 arm_cpu +enable_benchmark 0 # 是否开启benchmark， 默认是 0 +tipc_benchmark 0 # 是否开启tipc_benchmark，默认是 0 ``` -5. 启动调试，上述步骤完成后就可以使用ADB将文件夹 `debug/` push到手机上运行，步骤如下： +5. 执行预测命令 -```shell -# 执行编译，得到可执行文件clas_system -make -j - -# 将编译得到的可执行文件移动到debug文件夹中 -mv clas_system ./debug/ - -# 将上述debug文件夹push到手机上 -adb push debug /data/local/tmp/ - -adb shell -cd /data/local/tmp/debug -export LD_LIBRARY_PATH=/data/local/tmp/debug:$LD_LIBRARY_PATH +执行以下命令，可完成在手机上的预测。 -# clas_system可执行文件的使用方式为: -# ./clas_system 配置文件路径 测试图像路径 -./clas_system ./config.txt ./tabby_cat.jpg +```shell +adb shell 'export LD_LIBRARY_PATH=/data/local/tmp/arm_cpu/; /data/local/tmp/arm_cpu/clas_system /data/local/tmp/arm_cpu/config.txt /data/local/tmp/arm_cpu/tabby_cat.jpg' ``` -如果对代码做了修改，则需要重新编译并push到手机上。 - 运行效果如下：
@@ -263,3 +249,4 @@ A1：如果已经走通了上述步骤，更换模型只需要替换 `.nb` 模 Q2：换一个图测试怎么做？ A2：替换 debug 下的测试图像为你想要测试的图像，使用 ADB 再次 push 到手机上即可。 + diff --git a/deploy/lite_shitu/Makefile b/deploy/lite_shitu/Makefile index 64ee6aed81e09f9fc21f48aa1ab3ed4e81dd9b1c..df6685b6517d073d4c08bb5ec6dd34397fdbd072 100644 --- a/deploy/lite_shitu/Makefile +++ b/deploy/lite_shitu/Makefile @@ -17,7 +17,6 @@ ${info LITE_ROOT: $(abspath ${LITE_ROOT})} THIRD_PARTY_DIR=third_party ${info THIRD_PARTY_DIR: $(abspath ${THIRD_PARTY_DIR})} - OPENCV_VERSION=opencv4.1.0 OPENCV_LIBS = ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/libs/libopencv_imgcodecs.a \ ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/libs/libopencv_imgproc.a \ @@ -32,6 +31,8 @@ OPENCV_LIBS = ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/libs/libopencv_im ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/libtbb.a \ ${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/3rdparty/libs/libcpufeatures.a +FAISS_VERSION=faiss1.5.3 +FAISS_LIBS = ${THIRD_PARTY_DIR}/${FAISS_VERSION}/libs/${ARM_PLAT}/libfaiss.a LITE_LIBS = -L${LITE_ROOT}/cxx/lib/ -lpaddle_light_api_shared ############################################################### @@ -45,7 +46,7 @@ LITE_LIBS = -L${LITE_ROOT}/cxx/lib/ -lpaddle_light_api_shared # 2. Undo comment below line using `libpaddle_api_light_bundled.a` # LITE_LIBS = ${LITE_ROOT}/cxx/lib/libpaddle_api_light_bundled.a -CXX_LIBS = $(LITE_LIBS) ${OPENCV_LIBS} $(SYSTEM_LIBS) +CXX_LIBS = $(LITE_LIBS) ${OPENCV_LIBS} ${FAISS_LIBS} $(SYSTEM_LIBS) LOCAL_DIRSRCS=$(wildcard src/*.cc) LOCAL_SRCS=$(notdir $(LOCAL_DIRSRCS)) @@ -53,9 +54,17 @@ LOCAL_OBJS=$(patsubst %.cpp, %.o, $(patsubst %.cc, %.o, $(LOCAL_SRCS))) JSON_OBJS = json_reader.o json_value.o json_writer.o -pp_shitu: $(LOCAL_OBJS) $(JSON_OBJS) fetch_opencv +pp_shitu: $(LOCAL_OBJS) $(JSON_OBJS) fetch_opencv fetch_faiss $(CC) $(SYSROOT_LINK) $(CXXFLAGS_LINK) $(LOCAL_OBJS) $(JSON_OBJS) -o pp_shitu $(CXX_LIBS) $(LDFLAGS) +fetch_faiss: + @ test -d ${THIRD_PARTY_DIR} || mkdir ${THIRD_PARTY_DIR} + @ test -e ${THIRD_PARTY_DIR}/${FAISS_VERSION}.tar.gz || \ + (echo "fetch faiss libs" && \ + wget -P ${THIRD_PARTY_DIR} https://paddle-inference-dist.bj.bcebos.com/${FAISS_VERSION}.tar.gz) + @ test -d ${THIRD_PARTY_DIR}/${FAISS_VERSION} || \ + tar -xf ${THIRD_PARTY_DIR}/${FAISS_VERSION}.tar.gz -C ${THIRD_PARTY_DIR} + fetch_opencv: @ test -d ${THIRD_PARTY_DIR} || mkdir ${THIRD_PARTY_DIR} @ test -e ${THIRD_PARTY_DIR}/${OPENCV_VERSION}.tar.gz || \ @@ -74,11 +83,12 @@ fetch_json_code: LOCAL_INCLUDES = -I./ -Iinclude OPENCV_INCLUDE = -I${THIRD_PARTY_DIR}/${OPENCV_VERSION}/${ARM_PLAT}/include +FAISS_INCLUDE = -I${THIRD_PARTY_DIR}/${FAISS_VERSION}/include JSON_INCLUDE = -I${THIRD_PARTY_DIR}/jsoncpp_code/include -CXX_INCLUDES = ${LOCAL_INCLUDES} ${INCLUDES} ${OPENCV_INCLUDE} ${JSON_INCLUDE} -I$(LITE_ROOT)/cxx/include +CXX_INCLUDES = ${LOCAL_INCLUDES} ${INCLUDES} ${OPENCV_INCLUDE} ${FAISS_INCLUDE} ${JSON_INCLUDE} -I$(LITE_ROOT)/cxx/include -$(LOCAL_OBJS): %.o: src/%.cc fetch_opencv fetch_json_code +$(LOCAL_OBJS): %.o: src/%.cc fetch_opencv fetch_json_code fetch_faiss $(CC) $(SYSROOT_COMPLILE) $(CXX_DEFINES) $(CXX_INCLUDES) $(CXX_FLAGS) -c $< -o $@ $(JSON_OBJS): %.o: ${THIRD_PARTY_DIR}/jsoncpp_code/%.cpp fetch_json_code diff --git a/deploy/lite_shitu/README.md b/deploy/lite_shitu/README.md index 8f5462f69f87a0ac810fe7c7c7892b17520693c5..52871c3c16dc9990f9cf23de24b24cb54067cac6 100644 --- a/deploy/lite_shitu/README.md +++ b/deploy/lite_shitu/README.md @@ -2,7 +2,7 @@ 本教程将介绍基于[Paddle Lite](https://github.com/PaddlePaddle/Paddle-Lite) 在移动端部署PaddleClas PP-ShiTu模型的详细步骤。 -Paddle Lite是飞桨轻量化推理引擎，为手机、IOT端提供高效推理能力，并广泛整合跨平台硬件，为端侧部署及应用落地问题提供轻量化的部署方案。 +Paddle Lite是飞桨轻量化推理引擎，为手机、IoT端提供高效推理能力，并广泛整合跨平台硬件，为端侧部署及应用落地问题提供轻量化的部署方案。 ## 1. 准备环境 @@ -81,35 +81,134 @@ inference_lite_lib.android.armv8/ | `-- java Java 预测库demo ``` -## 2 开始运行 +## 2 模型准备 ### 2.1 模型准备 +PaddleClas 提供了转换并优化后的推理模型，可以直接参考下方 2.1.1 小节进行下载。如果需要使用其他模型，请参考后续 2.1.2 小节自行转换并优化模型。 -#### 2.1.1 模型准备 +#### 2.1.1 使用PaddleClas提供的推理模型 ```shell # 进入lite_ppshitu目录 cd $PaddleClas/deploy/lite_shitu -wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/lite/ppshitu_lite_models_v1.0.tar -tar -xf ppshitu_lite_models_v1.0.tar -rm -f ppshitu_lite_models_v1.0.tar +wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/lite/ppshitu_lite_models_v1.1.tar +tar -xf ppshitu_lite_models_v1.1.tar +rm -f ppshitu_lite_models_v1.1.tar ``` -#### 2.1.2将yaml文件转换成json文件 +#### 2.1.2 使用其他模型 + +Paddle-Lite 提供了多种策略来自动优化原始的模型，其中包括量化、子图融合、混合调度、Kernel优选等方法，使用Paddle-Lite的`opt`工具可以自动对inference模型进行优化，目前支持两种优化方式，优化后的模型更轻量，模型运行速度更快。 + +**注意**：如果已经准备好了 `.nb` 结尾的模型文件，可以跳过此步骤。 + +##### 2.1.2.1 安装paddle_lite_opt工具 + +安装`paddle_lite_opt`工具有如下两种方法： + +1. [**建议**]pip安装paddlelite并进行转换 + ```shell + pip install paddlelite==2.10rc + ``` + +2. 源码编译Paddle-Lite生成`paddle_lite_opt`工具 + + 模型优化需要Paddle-Lite的`opt`可执行文件，可以通过编译Paddle-Lite源码获得，编译步骤如下： + ```shell + # 如果准备环境时已经clone了Paddle-Lite，则不用重新clone Paddle-Lite + git clone https://github.com/PaddlePaddle/Paddle-Lite.git + cd Paddle-Lite + git checkout develop + # 启动编译 + ./lite/tools/build.sh build_optimize_tool + ``` + + 编译完成后，`opt`文件位于`build.opt/lite/api/`下，可通过如下方式查看`opt`的运行选项和使用方式； + ```shell + cd build.opt/lite/api/ + ./opt + ``` + + `opt`的使用方式与参数与上面的`paddle_lite_opt`完全一致。 + +之后使用`paddle_lite_opt`工具可以进行inference模型的转换。`paddle_lite_opt`的部分参数如下： + +|选项|说明| +|-|-| +|--model_file|待优化的PaddlePaddle模型（combined形式）的网络结构文件路径| +|--param_file|待优化的PaddlePaddle模型（combined形式）的权重文件路径| +|--optimize_out_type|输出模型类型，目前支持两种类型：protobuf和naive_buffer，其中naive_buffer是一种更轻量级的序列化/反序列化实现，默认为naive_buffer| +|--optimize_out|优化模型的输出路径| +|--valid_targets|指定模型可执行的backend，默认为arm。目前可支持x86、arm、opencl、npu、xpu，可以同时指定多个backend(以空格分隔)，Model Optimize Tool将会自动选择最佳方式。如果需要支持华为NPU（Kirin 810/990 Soc搭载的达芬奇架构NPU），应当设置为npu, arm| + +更详细的`paddle_lite_opt`工具使用说明请参考[使用opt转化模型文档](https://paddle-lite.readthedocs.io/zh/latest/user_guides/opt/opt_bin.html) + +`--model_file`表示inference模型的model文件地址，`--param_file`表示inference模型的param文件地址；`optimize_out`用于指定输出文件的名称（不需要添加`.nb`的后缀）。直接在命令行中运行`paddle_lite_opt`，也可以查看所有参数及其说明。 + + +##### 2.1.2.2 转换示例 + +下面介绍使用`paddle_lite_opt`完成主体检测模型和识别模型的预训练模型，转成inference模型，最终转换成Paddle-Lite的优化模型的过程。 + +1. 转换主体检测模型 + +```shell +# 当前目录为 $PaddleClas/deploy/lite_shitu +# $code_path需替换成相应的运行目录,可以根据需要，将$code_path设置成需要的目录 +export $code_path=~ +cd $code_path +git clone https://github.com/PaddlePaddle/PaddleDetection.git +# 进入PaddleDetection根目录 +cd PaddleDetection +# 将预训练模型导出为inference模型 +python tools/export_model.py -c configs/picodet/application/mainbody_detection/picodet_lcnet_x2_5_640_mainbody.yml -o weights=https://paddledet.bj.bcebos.com/models/picodet_lcnet_x2_5_640_mainbody.pdparams --output_dir=inference +# 将inference模型转化为Paddle-Lite优化模型 +paddle_lite_opt --model_file=inference/picodet_lcnet_x2_5_640_mainbody/model.pdmodel --param_file=inference/picodet_lcnet_x2_5_640_mainbody/model.pdiparams --optimize_out=inference/picodet_lcnet_x2_5_640_mainbody/mainbody_det +# 将转好的模型复制到lite_shitu目录下 +cd $PaddleClas/deploy/lite_shitu +mkdir models +cp $code_path/PaddleDetection/inference/picodet_lcnet_x2_5_640_mainbody/mainbody_det.nb $PaddleClas/deploy/lite_shitu/models +``` + +2. 转换识别模型 + +```shell +# 转换为Paddle-Lite模型 +paddle_lite_opt --model_file=inference/inference.pdmodel --param_file=inference/inference.pdiparams --optimize_out=inference/rec +# 将模型文件拷贝到lite_shitu下 +cp inference/rec.nb deploy/lite_shitu/models/ +cd deploy/lite_shitu +``` + +**注意**：`--optimize_out` 参数为优化后模型的保存路径，无需加后缀`.nb`；`--model_file` 参数为模型结构信息文件的路径，`--param_file` 参数为模型权重信息文件的路径，请注意文件名。 + +### 2.2 将yaml文件转换成json文件 ```shell # 如果测试单张图像 -python generate_json_config.py --det_model_path ppshitu_lite_models_v1.0/mainbody_PPLCNet_x2_5_640_quant_v1.0_lite.nb --rec_model_path ppshitu_lite_models_v1.0/general_PPLCNet_x2_5_quant_v1.0_lite.nb --rec_label_path ppshitu_lite_models_v1.0/label.txt --img_path images/demo.jpg +python generate_json_config.py --det_model_path ppshitu_lite_models_v1.1/mainbody_PPLCNet_x2_5_640_quant_v1.1_lite.nb --rec_model_path ppshitu_lite_models_v1.1/general_PPLCNet_x2_5_lite_v1.1_infer.nb --img_path images/demo.jpg # or # 如果测试多张图像 -python generate_json_config.py --det_model_path ppshitu_lite_models_v1.0/mainbody_PPLCNet_x2_5_640_quant_v1.0_lite.nb --rec_model_path ppshitu_lite_models_v1.0/general_PPLCNet_x2_5_quant_v1.0_lite.nb --rec_label_path ppshitu_lite_models_v1.0/label.txt --img_dir images - +python generate_json_config.py --det_model_path ppshitu_lite_models_v1.1/mainbody_PPLCNet_x2_5_640_quant_v1.1_lite.nb --rec_model_path ppshitu_lite_models_v1.1/general_PPLCNet_x2_5_lite_v1.1_infer.nb --img_dir images # 执行完成后，会在lit_shitu下生成shitu_config.json配置文件 +``` + +### 2.3 index字典转换 +由于python的检索库字典，使用`pickle`进行的序列化存储，导致C++不方便读取，因此需要进行转换 + +```shell +# 下载瓶装饮料数据集 +wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/data/drink_dataset_v1.0.tar && tar -xf drink_dataset_v1.0.tar +rm -rf drink_dataset_v1.0.tar +# 转化id_map.pkl为id_map.txt +python transform_id_map.py -c ../configs/inference_drink.yaml ``` +转换成功后，会在`IndexProcess.index_dir`目录下生成`id_map.txt`。 + -### 2.2 与手机联调 +### 2.4 与手机联调 首先需要进行一些准备工作。 1. 准备一台arm8的安卓手机，如果编译的预测库是armv7，则需要arm7的手机，并修改Makefile中`ARM_ABI=arm7`。 @@ -160,7 +259,8 @@ make ARM_ABI=arm8 ```shell mkdir deploy -mv ppshitu_lite_models_v1.0 deploy/ +mv ppshitu_lite_models_v1.1 deploy/ +mv drink_dataset_v1.0 deploy/ mv images deploy/ mv shitu_config.json deploy/ cp pp_shitu deploy/ @@ -173,13 +273,13 @@ cp ../../../cxx/lib/libpaddle_light_api_shared.so deploy/ ```shell deploy/ -|-- ppshitu_lite_models_v1.0/ -| |--mainbody_PPLCNet_x2_5_640_v1.0_lite.nb 优化后的主体检测模型文件 -| |--general_PPLCNet_x2_5_quant_v1.0_lite.nb 优化后的识别模型文件 -| |--label.txt 识别模型的label文件 +|-- ppshitu_lite_models_v1.1/ +| |--mainbody_PPLCNet_x2_5_640_quant_v1.1_lite.nb 优化后的主体检测模型文件 +| |--general_PPLCNet_x2_5_lite_v1.1_infer.nb 优化后的识别模型文件 |-- images/ | |--demo.jpg 图片文件 -| ... 图片文件 +|-- drink_dataset_v1.0/ 瓶装饮料demo数据 +| |--index 检索index目录 |-- pp_shitu 生成的移动端执行文件 |-- shitu_config.json 执行时参数配置文件 |-- libpaddle_light_api_shared.so Paddle-Lite库文件 @@ -207,8 +307,10 @@ chmod 777 pp_shitu 如果对代码做了修改，则需要重新编译并push到手机上。 运行效果如下： - -![](../../docs/images/ppshitu_lite_demo.png) +``` +images/demo.jpg: + result0: bbox[253, 275, 1146, 872], score: 0.974196, label: 伊藤园_果蔬汁 +``` ## FAQ Q1：如果想更换模型怎么办，需要重新按照流程走一遍吗？ diff --git a/deploy/lite_shitu/generate_json_config.py b/deploy/lite_shitu/generate_json_config.py index 1525cdab9e223eb480dec2103328a22c91551f65..642dfcd9d6a46e2894ec0f01f0914a5347bc8d72 100644 --- a/deploy/lite_shitu/generate_json_config.py +++ b/deploy/lite_shitu/generate_json_config.py @@ -95,7 +95,7 @@ def main(): config_json["Global"]["det_model_path"] = args.det_model_path config_json["Global"]["rec_model_path"] = args.rec_model_path config_json["Global"]["rec_label_path"] = args.rec_label_path - config_json["Global"]["label_list"] = config_yaml["Global"]["labe_list"] + config_json["Global"]["label_list"] = config_yaml["Global"]["label_list"] config_json["Global"]["rec_nms_thresold"] = config_yaml["Global"][ "rec_nms_thresold"] config_json["Global"]["max_det_results"] = config_yaml["Global"][ @@ -130,6 +130,8 @@ def main(): y["type"] = k config_json["RecPreProcess"]["transform_ops"].append(y) + # set IndexProces + config_json["IndexProcess"] = config_yaml["IndexProcess"] with open('shitu_config.json', 'w') as fd: json.dump(config_json, fd, indent=4) diff --git a/deploy/lite_shitu/include/recognition.h b/deploy/lite_shitu/include/feature_extractor.h similarity index 72% rename from deploy/lite_shitu/include/recognition.h rename to deploy/lite_shitu/include/feature_extractor.h index 0c45e946e8b88d161fcbe470ccefc6e370cf8b1e..1961459ecfab149695890df60cef550ed5177b52 100644 --- a/deploy/lite_shitu/include/recognition.h +++ b/deploy/lite_shitu/include/feature_extractor.h @@ -36,10 +36,9 @@ struct RESULT { float score; }; -class Recognition { - +class FeatureExtract { public: - explicit Recognition(const Json::Value &config_file) { + explicit FeatureExtract(const Json::Value &config_file) { MobileConfig config; if (config_file["Global"]["rec_model_path"].as().empty()) { std::cout << "Please set [rec_model_path] in config file" << std::endl; @@ -53,29 +52,8 @@ public: std::cout << "Please set [rec_label_path] in config file" << std::endl; exit(-1); } - LoadLabel(config_file["Global"]["rec_label_path"].as()); SetPreProcessParam(config_file["RecPreProcess"]["transform_ops"]); - if (!config_file["Global"].isMember("return_k")){ - this->topk = config_file["Global"]["return_k"].as(); - } - printf("rec model create!\n"); - } - - void LoadLabel(std::string path) { - std::ifstream file; - std::vector label_list; - file.open(path); - while (file) { - std::string line; - std::getline(file, line); - std::string::size_type pos = line.find(" "); - if (pos != std::string::npos) { - line = line.substr(pos); - } - this->label_list.push_back(line); - } - file.clear(); - file.close(); + printf("feature extract model create!\n"); } void SetPreProcessParam(const Json::Value &config_file) { @@ -97,19 +75,17 @@ public: } } - std::vector RunRecModel(const cv::Mat &img, double &cost_time); - std::vector PostProcess(const float *output_data, int output_size, - cv::Mat &output_image); + void RunRecModel(const cv::Mat &img, double &cost_time, std::vector &feature); + //void PostProcess(std::vector &feature); cv::Mat ResizeImage(const cv::Mat &img); void NeonMeanScale(const float *din, float *dout, int size); private: std::shared_ptr predictor; - std::vector label_list; + //std::vector label_list; std::vector mean = {0.485f, 0.456f, 0.406f}; std::vector std = {1 / 0.229f, 1 / 0.224f, 1 / 0.225f}; double scale = 0.00392157; float size = 224; - int topk = 5; }; } // namespace PPShiTu diff --git a/deploy/lite_shitu/include/utils.h b/deploy/lite_shitu/include/utils.h index 18a04cf344327c2a9c363ac73c8aff5d26f15eef..a3b57c882561577defff97e384fb775b78204f36 100644 --- a/deploy/lite_shitu/include/utils.h +++ b/deploy/lite_shitu/include/utils.h @@ -16,7 +16,7 @@ #include #include -#include +#include #include #include #include diff --git a/deploy/lite_shitu/include/vector_search.h b/deploy/lite_shitu/include/vector_search.h new file mode 100644 index 0000000000000000000000000000000000000000..89ef7733ab86c534a5c507cb4f87c9d4597dba15 --- /dev/null +++ b/deploy/lite_shitu/include/vector_search.h @@ -0,0 +1,73 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once +#ifdef WIN32 +#define OS_PATH_SEP "\\" +#else +#define OS_PATH_SEP "/" +#endif + +#include "json/json.h" +#include +#include +#include +#include + +namespace PPShiTu { +struct SearchResult { + std::vector I; + std::vector D; + int return_k; +}; + +class VectorSearch { +public: + explicit VectorSearch(const Json::Value &config) { + // IndexProcess + this->index_dir = config["IndexProcess"]["index_dir"].as(); + this->return_k = config["IndexProcess"]["return_k"].as(); + this->score_thres = config["IndexProcess"]["score_thres"].as(); + this->max_query_number = config["Global"]["max_det_results"].as() + 1; + + LoadIdMap(); + LoadIndexFile(); + this->I.resize(this->return_k * this->max_query_number); + this->D.resize(this->return_k * this->max_query_number); + printf("faiss index load success!\n"); + }; + + void LoadIdMap(); + + void LoadIndexFile(); + + const SearchResult &Search(float *feature, int query_number); + + const std::string &GetLabel(faiss::Index::idx_t ind); + + const float &GetThreshold() { return this->score_thres; } + +private: + std::string index_dir; + int return_k = 5; + float score_thres = 0.5; + + std::map id_map; + faiss::Index *index; + int max_query_number = 6; + std::vector D; + std::vector I; + SearchResult sr; +}; +} diff --git a/deploy/lite_shitu/src/recognition.cc b/deploy/lite_shitu/src/feature_extractor.cc similarity index 65% rename from deploy/lite_shitu/src/recognition.cc rename to deploy/lite_shitu/src/feature_extractor.cc index 0e711f386b31556e703bd89839e755ce07d88e22..aca5c1cbbe5c70cd214c922609831e9350be28a0 100644 --- a/deploy/lite_shitu/src/recognition.cc +++ b/deploy/lite_shitu/src/feature_extractor.cc @@ -12,12 +12,12 @@ // See the License for the specific language governing permissions and // limitations under the License. -#include "include/recognition.h" +#include "include/feature_extractor.h" namespace PPShiTu { -std::vector Recognition::RunRecModel(const cv::Mat &img, - double &cost_time) { - +void FeatureExtract::RunRecModel(const cv::Mat &img, + double &cost_time, + std::vector &feature) { // Read img cv::Mat resize_image = ResizeImage(img); @@ -38,8 +38,7 @@ std::vector Recognition::RunRecModel(const cv::Mat &img, // Get output and post process std::unique_ptr output_tensor( - std::move(this->predictor->GetOutput(1))); - auto *output_data = output_tensor->data(); + std::move(this->predictor->GetOutput(0))); //only one output auto end = std::chrono::system_clock::now(); auto duration = std::chrono::duration_cast(end - start); @@ -47,17 +46,27 @@ std::vector Recognition::RunRecModel(const cv::Mat &img, std::chrono::microseconds::period::num / std::chrono::microseconds::period::den; + //do postprocess int output_size = 1; for (auto dim : output_tensor->shape()) { output_size *= dim; } + feature.resize(output_size); + output_tensor->CopyToCpu(feature.data()); - cv::Mat output_image; - auto results = PostProcess(output_data, output_size, output_image); - return results; + //postprocess include sqrt or binarize. + //PostProcess(feature); + return; } -void Recognition::NeonMeanScale(const float *din, float *dout, int size) { +// void FeatureExtract::PostProcess(std::vector &feature){ +// float feature_sqrt = std::sqrt(std::inner_product( +// feature.begin(), feature.end(), feature.begin(), 0.0f)); +// for (int i = 0; i < feature.size(); ++i) +// feature[i] /= feature_sqrt; +// } + +void FeatureExtract::NeonMeanScale(const float *din, float *dout, int size) { if (this->mean.size() != 3 || this->std.size() != 3) { std::cerr << "[ERROR] mean or scale size must equal to 3\n"; @@ -99,45 +108,9 @@ void Recognition::NeonMeanScale(const float *din, float *dout, int size) { } } -cv::Mat Recognition::ResizeImage(const cv::Mat &img) { +cv::Mat FeatureExtract::ResizeImage(const cv::Mat &img) { cv::Mat resize_img; cv::resize(img, resize_img, cv::Size(this->size, this->size)); return resize_img; } -std::vector Recognition::PostProcess(const float *output_data, - int output_size, - cv::Mat &output_image) { - - int max_indices[this->topk]; - double max_scores[this->topk]; - for (int i = 0; i < this->topk; i++) { - max_indices[i] = 0; - max_scores[i] = 0; - } - for (int i = 0; i < output_size; i++) { - float score = output_data[i]; - int index = i; - for (int j = 0; j < this->topk; j++) { - if (score > max_scores[j]) { - index += max_indices[j]; - max_indices[j] = index - max_indices[j]; - index -= max_indices[j]; - score += max_scores[j]; - max_scores[j] = score - max_scores[j]; - score -= max_scores[j]; - } - } - } - - std::vector results(this->topk); - for (int i = 0; i < results.size(); i++) { - results[i].class_name = "Unknown"; - if (max_indices[i] >= 0 && max_indices[i] < this->label_list.size()) { - results[i].class_name = this->label_list[max_indices[i]]; - } - results[i].score = max_scores[i]; - results[i].class_id = max_indices[i]; - } - return results; -} } diff --git a/deploy/lite_shitu/src/main.cc b/deploy/lite_shitu/src/main.cc index 690162a4939434270a2d8d7d6c23e02cf35478a4..3f278dc778701a7a7591e74336e0f86fe52105ea 100644 --- a/deploy/lite_shitu/src/main.cc +++ b/deploy/lite_shitu/src/main.cc @@ -24,9 +24,10 @@ #include #include "include/config_parser.h" +#include "include/feature_extractor.h" #include "include/object_detector.h" #include "include/preprocess_op.h" -#include "include/recognition.h" +#include "include/vector_search.h" #include "json/json.h" Json::Value RT_Config; @@ -111,14 +112,18 @@ void DetPredictImage(const std::vector &batch_imgs, } } -void PrintResult(const std::string &image_path, - std::vector &det_result) { - printf("%s:\n", image_path.c_str()); +void PrintResult(std::string &img_path, + std::vector &det_result, + PPShiTu::VectorSearch &vector_search, + PPShiTu::SearchResult &search_result) { + printf("%s:\n", img_path.c_str()); for (int i = 0; i < det_result.size(); ++i) { + int t = i; printf("\tresult%d: bbox[%d, %d, %d, %d], score: %f, label: %s\n", i, - det_result[i].rect[0], det_result[i].rect[1], det_result[i].rect[2], - det_result[i].rect[3], det_result[i].rec_result[0].score, - det_result[i].rec_result[0].class_name.c_str()); + det_result[t].rect[0], det_result[t].rect[1], det_result[t].rect[2], + det_result[t].rect[3], det_result[t].confidence, + vector_search.GetLabel(search_result.I[search_result.return_k * t]) + .c_str()); } } @@ -159,11 +164,16 @@ int main(int argc, char **argv) { RT_Config["Global"]["cpu_num_threads"].as(), RT_Config["Global"]["batch_size"].as()); // create rec model - PPShiTu::Recognition rec(RT_Config); + PPShiTu::FeatureExtract rec(RT_Config); + PPShiTu::VectorSearch searcher(RT_Config); // Do inference on input image std::vector det_result; std::vector batch_imgs; + + // for vector search + std::vector feature; + std::vector features; double rec_time; if (!RT_Config["Global"]["infer_imgs"].as().empty() || !img_dir.empty()) { @@ -178,8 +188,7 @@ int main(int argc, char **argv) { return -1; } } else { - cv::glob(img_dir, - cv_all_img_paths); + cv::glob(img_dir, cv_all_img_paths); for (const auto &img_path : cv_all_img_paths) { all_img_paths.push_back(img_path); } @@ -199,24 +208,25 @@ int main(int argc, char **argv) { RT_Config["Global"]["max_det_results"].as(), false, &det); // add the whole image for recognition to improve recall - PPShiTu::ObjectResult result_whole_img = { - {0, 0, srcimg.cols, srcimg.rows}, 0, 1.0}; - det_result.push_back(result_whole_img); +// PPShiTu::ObjectResult result_whole_img = { +// {0, 0, srcimg.cols, srcimg.rows}, 0, 1.0}; +// det_result.push_back(result_whole_img); // get rec result + PPShiTu::SearchResult search_result; for (int j = 0; j < det_result.size(); ++j) { int w = det_result[j].rect[2] - det_result[j].rect[0]; int h = det_result[j].rect[3] - det_result[j].rect[1]; cv::Rect rect(det_result[j].rect[0], det_result[j].rect[1], w, h); cv::Mat crop_img = srcimg(rect); - std::vector result = - rec.RunRecModel(crop_img, rec_time); - det_result[j].rec_result.assign(result.begin(), result.end()); + rec.RunRecModel(crop_img, rec_time, feature); + features.insert(features.end(), feature.begin(), feature.end()); } - // rec nms - PPShiTu::nms(det_result, - RT_Config["Global"]["rec_nms_thresold"].as(), true); - PrintResult(img_path, det_result); + + // do vectore search + search_result = searcher.Search(features.data(), det_result.size()); + PrintResult(img_path, det_result, searcher, search_result); + batch_imgs.clear(); det_result.clear(); } diff --git a/deploy/lite_shitu/src/vector_search.cc b/deploy/lite_shitu/src/vector_search.cc new file mode 100644 index 0000000000000000000000000000000000000000..ea848959b651eb04effc25ad9efb7eb497ef2025 --- /dev/null +++ b/deploy/lite_shitu/src/vector_search.cc @@ -0,0 +1,67 @@ +// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "include/vector_search.h" +#include +#include +#include +#include +#include + +namespace PPShiTu { +// load the vector.index +void VectorSearch::LoadIndexFile() { + std::string file_path = this->index_dir + OS_PATH_SEP + "vector.index"; + const char *fname = file_path.c_str(); + this->index = faiss::read_index(fname, 0); +} + +// load id_map.txt +void VectorSearch::LoadIdMap() { + std::string file_path = this->index_dir + OS_PATH_SEP + "id_map.txt"; + std::ifstream in(file_path); + std::string line; + std::vector m_vec; + if (in) { + while (getline(in, line)) { + std::regex ws_re("\\s+"); + std::vector v( + std::sregex_token_iterator(line.begin(), line.end(), ws_re, -1), + std::sregex_token_iterator()); + if (v.size() != 2) { + std::cout << "The number of element for each line in : " << file_path + << "must be 2, exit the program..." << std::endl; + exit(1); + } else + this->id_map.insert(std::pair( + std::stol(v[0], nullptr, 10), v[1])); + } + } +} + +// doing search +const SearchResult &VectorSearch::Search(float *feature, int query_number) { + this->D.resize(this->return_k * query_number); + this->I.resize(this->return_k * query_number); + this->index->search(query_number, feature, return_k, D.data(), I.data()); + this->sr.return_k = this->return_k; + this->sr.D = this->D; + this->sr.I = this->I; + return this->sr; +} + +const std::string &VectorSearch::GetLabel(faiss::Index::idx_t ind) { + return this->id_map.at(ind); +} +} \ No newline at end of file diff --git a/deploy/lite_shitu/transform_id_map.py b/deploy/lite_shitu/transform_id_map.py new file mode 100644 index 0000000000000000000000000000000000000000..3ad6b6f9bdb15d0feb746e8f39ea88a5243836ae --- /dev/null +++ b/deploy/lite_shitu/transform_id_map.py @@ -0,0 +1,39 @@ +import argparse +import os +import pickle + +import yaml + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument('-c', '--config', type=str, required=True) + args = parser.parse_args() + return args + + +def main(): + args = parse_args() + with open(args.config) as fd: + config = yaml.load(fd.read(), yaml.FullLoader) + index_dir = "" + try: + index_dir = config["IndexProcess"]["index_dir"] + except Exception as e: + print("The IndexProcess.index_dir in config_file dose not exist") + exit(1) + id_map_path = os.path.join(index_dir, "id_map.pkl") + assert os.path.exists( + id_map_path), "The id_map file dose not exist: {}".format(id_map_path) + + with open(id_map_path, "rb") as fd: + ids = pickle.load(fd) + with open(os.path.join(index_dir, "id_map.txt"), "w") as fd: + for k, v in ids.items(): + v = v.split("\t")[1] + fd.write(str(k) + " " + v + "\n") + print('Transform id_map sucess') + + +if __name__ == "__main__": + main() diff --git a/deploy/python/predict_det.py b/deploy/python/predict_det.py index e4e0a24a6dbc6c62f82810c865096f768ebd182b..37a7bf5018c3b5dc78e897b532303f70b0d3957d 100644 --- a/deploy/python/predict_det.py +++ b/deploy/python/predict_det.py @@ -128,13 +128,10 @@ class DetPredictor(Predictor): results = [] if reduce(lambda x, y: x * y, np_boxes.shape) < 6: print('[WARNNING] No object detected.') - results = np.array([]) else: - results = np_boxes - - results = self.parse_det_results(results, - self.config["Global"]["threshold"], - self.config["Global"]["labe_list"]) + results = self.parse_det_results( + np_boxes, self.config["Global"]["threshold"], + self.config["Global"]["label_list"]) return results diff --git a/deploy/utils/predictor.py b/deploy/utils/predictor.py index d44ae03df57def97bed6ab91205c327c3d318a2d..7fd1d6dccb61b86f1fece2e3a909c7005f93ca8a 100644 --- a/deploy/utils/predictor.py +++ b/deploy/utils/predictor.py @@ -11,6 +11,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +import platform import os import argparse import base64 @@ -50,8 +51,10 @@ class Predictor(object): else: config.disable_gpu() if args.enable_mkldnn: - # cache 10 different shapes for mkldnn to avoid memory leak - config.set_mkldnn_cache_capacity(10) + # there is no set_mkldnn_cache_capatity() on macOS + if platform.system() != "Darwin": + # cache 10 different shapes for mkldnn to avoid memory leak + config.set_mkldnn_cache_capacity(10) config.enable_mkldnn() config.set_cpu_math_library_num_threads(args.cpu_num_threads) diff --git a/docs/en/algorithm_introduction/ISE_ReID_en.md b/docs/en/algorithm_introduction/ISE_ReID_en.md new file mode 100644 index 0000000000000000000000000000000000000000..e509a52018625fa1ef49f7495af949ad68d30909 --- /dev/null +++ b/docs/en/algorithm_introduction/ISE_ReID_en.md @@ -0,0 +1,60 @@ +# ISE +--- +## Catalogue + +- [1. Introduction](#1) +- [2. Performance on Market1501 and MSMT17](#2) +- [3. Test](#3) +- [4. Reference](#4) + + +## 1. Introduction + +ISE (Implicit Sample Extension) is a simple, efficient, and effective learning algorithm for unsupervised person Re-ID. ISE generates what we call support samples around the cluster boundaries. The sample generation process in ISE depends on two critical mechanisms, i.e., a progressive linear interpolation strategy and a label-preserving loss function. The generated support samples from ISE provide complementary information, which can nicely handle the "sub and mixed" clustering errors. ISE achieves superior performance than other unsupervised methods on Market1501 and MSMT17 datasets. + +> [**Implicit Sample Extension for Unsupervised Person Re-Identification**](https://arxiv.org/abs/2204.06892v1)
+> Xinyu Zhang, Dongdong Li, Zhigang Wang, Jian Wang, Errui Ding, Javen Qinfeng Shi, Zhaoxiang Zhang, Jingdong Wang
+> CVPR2022 + +![image](../../images/ISE_ReID/ISE_pipeline.png) + + +## 2. Performance on Market1501 and MSMT17 + +The main results on Market1501 (M) and MSMT17 (MS). PIL denotes the progressive linear interpolation strategy. LP represents the label-preserving loss function. + +| Methods | M | Link | MS | Link | +| --- | -- | -- | -- | - | +| Baseline | 82.5 (92.5) | - | 30.1 (58.6) | - | +| ISE (+PIL) | 83.9 (93.9) | - | 33.5 (63.9) | - | +| ISE (+LP) | 83.6 (92.7) | - | 31.4 (59.9) | - | +| ISE (Ours) (+PIL+LP) | **84.7 (94.0)** | [ISE_M](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ISE_M_model.pdparams) | **35.0 (64.7)** | [ISE_MS](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ISE_MS_model.pdparams) | + + +## 3. Test +The training code is coming soon. We first release the test code with the pretrained models. + +**Test:** You can simply run the following script for the evaluation. + +``` +python tools/eval.py -c ./ppcls/configs/Person/ResNet50_UReID_infer.yaml +``` +**Steps:** +1. Download the pretrained model first, and put the model into: ```./pd_model_trace/ISE/```. +2. Change the dataset name in: ```./ppcls/configs/Person/ResNet50_UReID_infer.yaml```. +3. Run the above script. + + + +## 4. Reference + +If you find ISE useful in your research, please kindly consider citing our paper: + +``` +@inproceedings{zhang2022Implicit, + title={Implicit Sample Extension for Unsupervised Person Re-Identification}, + author={Xinyu Zhang, Dongdong Li, Zhigang Wang, Jian Wang, Errui Ding, Javen Qinfeng Shi, Zhaoxiang Zhang, Jingdong Wang}, + booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, + year={2022} +} +``` diff --git a/docs/en/algorithm_introduction/ImageNet_models_en.md b/docs/en/algorithm_introduction/ImageNet_models_en.md index 580aeaa1ebfc57b15e680c688a41242083353afa..b78061267b65d44a5c175368a2575bc2fc277f50 100644 --- a/docs/en/algorithm_introduction/ImageNet_models_en.md +++ b/docs/en/algorithm_introduction/ImageNet_models_en.md @@ -27,7 +27,11 @@ - [20. DLA series](#20) - [21. RedNet series](#21) - [22. TNT series](#22) -- [23. Other models](#23) +- [23. CSwinTransformer series](#23) +- [24. PVTV2 series](#24) +- [25. MobileViT series](#25) +- [26. Other models](#26) +- [Reference](#reference) @@ -63,7 +67,7 @@ Accuracy and inference time of the prtrained models based on SSLD distillation a | Model | Top-1 Acc | Reference
Top-1 Acc | Acc gain | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | |---------------------|-----------|-----------|---------------|----------------|-----------|----------|-----------|-----------------------------------|-----------------------------------|-----------------------------------| -| ResNet34_vd_ssld | 0.797 | 0.760 | 0.037 | 2.00 | 3.28 | 5.84 | 3.93 | 21.84 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_ssld_pretrained.pdparams)   | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet34_vd_ssld.tar)   | +| ResNet34_vd_ssld | 0.797 | 0.760 | 0.037 | 2.00 | 3.28 | 5.84 | 3.93 | 21.84 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_ssld_pretrained.pdparams)   | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet34_vd_ssld_infer.tar)   | | ResNet50_vd_ssld | 0.830 | 0.792 | 0.039 | 2.60 | 4.86 | 7.63 | 4.35 | 25.63 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_ssld_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet50_vd_ssld_infer.tar) | | ResNet101_vd_ssld | 0.837 | 0.802 | 0.035 | 4.43 | 8.25 | 12.60 | 8.08 | 44.67 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_vd_ssld_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet101_vd_ssld_infer.tar) | | Res2Net50_vd_26w_4s_ssld | 0.831 | 0.798 | 0.033 | 3.59 | 6.35 | 9.50 | 4.28 | 25.76 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_ssld_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/Res2Net50_vd_26w_4s_ssld_infer.tar) | @@ -101,7 +105,7 @@ Accuracy and inference time of the prtrained models based on SSLD distillation a -## 3. PP-LCNet series +## 3. PP-LCNet series [[28](#ref28)] The accuracy and speed indicators of the PP-LCNet series models are shown in the following table. For more information about this series of models, please refer to: [PP-LCNet series model documents](../models/PP-LCNet_en.md)。 @@ -118,7 +122,7 @@ The accuracy and speed indicators of the PP-LCNet series models are shown in the -## 4. ResNet series +## 4. ResNet series [[1](#ref1)] The accuracy and speed indicators of ResNet and ResNet_vd series models are shown in the following table. For more information about this series of models, please refer to: [ResNet and ResNet_vd series model documents](../models/ResNet_and_vd_en.md)。 @@ -142,7 +146,7 @@ The accuracy and speed indicators of ResNet and ResNet_vd series models are show -## 5. Mobile series +## 5. Mobile series [[3](#ref3)][[4](#ref4)][[5](#ref5)][[6](#ref6)][[23](#ref23)] The accuracy and speed indicators of the mobile series models are shown in the following table. For more information about this series, please refer to: [Mobile series model documents](../models/Mobile_en.md)。 @@ -191,7 +195,7 @@ The accuracy and speed indicators of the mobile series models are shown in the f -## 6. SEResNeXt and Res2Net series +## 6. SEResNeXt and Res2Net series [[7](#ref7)][[8](#ref8)][[9](#ref9)] The accuracy and speed indicators of the SEResNeXt and Res2Net series models are shown in the following table. For more information about the models of this series, please refer to: [SEResNeXt and Res2Net series model documents](../models/SEResNext_and_Res2Net_en.md). @@ -226,7 +230,7 @@ The accuracy and speed indicators of the SEResNeXt and Res2Net series models are -## 7. DPN and DenseNet series +## 7. DPN and DenseNet series [[14](#ref14)][[15](#ref15)] The accuracy and speed indicators of the DPN and DenseNet series models are shown in the following table. For more information about the models of this series, please refer to: [DPN and DenseNet series model documents](../models/DPN_DenseNet_en.md). @@ -244,11 +248,9 @@ The accuracy and speed indicators of the DPN and DenseNet series models are show | DPN107 | 0.8089 | 0.9532 | 19.46 | 35.62 | 50.22 | 18.38 | 87.13 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN107_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DPN107_infer.tar) | | DPN131 | 0.8070 | 0.9514 | 19.64 | 34.60 | 47.42 | 16.09 | 79.48 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN131_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DPN131_infer.tar) | - - -## 8. HRNet series +## 8. HRNet series [[13](#ref13)] The accuracy and speed indicators of the HRNet series models are shown in the following table. For more information about the models of this series, please refer to: [HRNet series model documents](../models/HRNet_en.md). @@ -268,7 +270,7 @@ The accuracy and speed indicators of the HRNet series models are shown in the fo -## 9. Inception series +## 9. Inception series [[10](#ref10)][[11](#ref11)][[12](#ref12)][[26](#ref26)] The accuracy and speed indicators of the Inception series models are shown in the following table. For more information about this series of models, please refer to: [Inception series model documents](../models/Inception_en.md). @@ -285,7 +287,7 @@ The accuracy and speed indicators of the Inception series models are shown in th -## 10. EfficientNet and ResNeXt101_wsl series +## 10. EfficientNet and ResNeXt101_wsl series [[16](#ref16)][[17](#ref17)] The accuracy and speed indicators of the EfficientNet and ResNeXt101_wsl series models are shown in the following table. For more information about this series of models, please refer to: [EfficientNet and ResNeXt101_wsl series model documents](../models/EfficientNet_and_ResNeXt101_wsl_en.md). @@ -308,7 +310,7 @@ The accuracy and speed indicators of the EfficientNet and ResNeXt101_wsl series -## 11. ResNeSt and RegNet series +## 11. ResNeSt and RegNet series [[24](#ref24)][[25](#ref25)] The accuracy and speed indicators of the ResNeSt and RegNet series models are shown in the following table. For more information about the models of this series, please refer to: [ResNeSt and RegNet series model documents](../models/ResNeSt_RegNet_en.md). @@ -320,11 +322,10 @@ The accuracy and speed indicators of the ResNeSt and RegNet series models are sh -## 12. ViT and DeiT series +## 12. ViT and DeiT series [[31](#ref31)][[32](#ref32)] The accuracy and speed indicators of ViT (Vision Transformer) and DeiT (Data-efficient Image Transformers) series models are shown in the following table. For more information about this series of models, please refer to: [ViT_and_DeiT series model documents](../models/ViT_and_DeiT_en.md). - | Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | |------------------------|-----------|-----------|------------------|------------------|----------|------------------------|------------------------|------------------------|------------------------| | ViT_small_
patch16_224 | 0.7769 | 0.9342 | 3.71 | 9.05 | 16.72 | 9.41 | 48.60 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_small_patch16_224_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ViT_small_patch16_224_infer.tar) | @@ -335,8 +336,6 @@ The accuracy and speed indicators of ViT (Vision Transformer) and DeiT (Data-eff |ViT_large_
patch16_384| 0.8513 | 0.9736 | 39.51 | 152.46 | 304.06 | 174.70 | 304.12 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_384_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ViT_large_patch16_384_infer.tar) | |ViT_large_
patch32_384| 0.8153 | 0.9608 | 11.44 | 36.09 | 70.63 | 44.24 | 306.48 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch32_384_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ViT_large_patch32_384_infer.tar) | - - | Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | |------------------------|-----------|-----------|------------------|------------------|----------|------------------------|------------------------|------------------------|------------------------| | DeiT_tiny_
patch16_224 | 0.718 | 0.910 | 3.61 | 3.94 | 6.10 | 1.07 | 5.68 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_patch16_224_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DeiT_tiny_patch16_224_infer.tar) | @@ -350,11 +349,10 @@ The accuracy and speed indicators of ViT (Vision Transformer) and DeiT (Data-eff -## 13. RepVGG series +## 13. RepVGG series [[36](#ref36)] The accuracy and speed indicators of RepVGG series models are shown in the following table. For more introduction, please refer to: [RepVGG series model documents](../models/RepVGG_en.md). - | Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | |------------------------|-----------|-----------|------------------|------------------|----------|-----------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------| | RepVGG_A0 | 0.7131 | 0.9016 | | | | 1.36 | 8.31 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A0_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/RepVGG_A0_infer.tar) | @@ -370,7 +368,7 @@ The accuracy and speed indicators of RepVGG series models are shown in the follo -## 14. MixNet series +## 14. MixNet series [[29](#ref29)] The accuracy and speed indicators of the MixNet series models are shown in the following table. For more introduction, please refer to: [MixNet series model documents](../models/MixNet_en.md). @@ -382,7 +380,7 @@ The accuracy and speed indicators of the MixNet series models are shown in the f -## 15. ReXNet series +## 15. ReXNet series [[30](#ref30)] The accuracy and speed indicators of ReXNet series models are shown in the following table. For more introduction, please refer to: [ReXNet series model documents](../models/ReXNet_en.md). @@ -396,7 +394,7 @@ The accuracy and speed indicators of ReXNet series models are shown in the follo -## 16. SwinTransformer series +## 16. SwinTransformer series [[27](#ref27)] The accuracy and speed indicators of SwinTransformer series models are shown in the following table. For more introduction, please refer to: [SwinTransformer series model documents](../models/SwinTransformer_en.md). @@ -408,20 +406,20 @@ The accuracy and speed indicators of SwinTransformer series models are shown in | SwinTransformer_base_patch4_window12_384 | 0.8439 | 0.9693 | 19.52 | 64.56 | 123.30 | 44.45 | 87.70 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_base_patch4_window12_384_infer.tar) | | SwinTransformer_base_patch4_window7_224[1] | 0.8487 | 0.9746 | 13.53 | 23.46 | 39.13 | 15.13 | 87.70 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_22kto1k_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_base_patch4_window7_224_infer.tar) | | SwinTransformer_base_patch4_window12_384[1] | 0.8642 | 0.9807 | 19.65 | 64.72 | 123.42 | 44.45 | 87.70 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_22kto1k_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_base_patch4_window12_384_infer.tar) | -| SwinTransformer_large_patch4_window7_224[1] | 0.8596 | 0.9783 | 15.74 | 38.57 | 71.49 | 34.02 | 196.43 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window7_224_infer.tar) | -| SwinTransformer_large_patch4_window12_384[1] | 0.8719 | 0.9823 | 32.61 | 116.59 | 223.23 | 99.97 | 196.43 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window12_384_infer.tar) | +| SwinTransformer_large_patch4_window7_224[1] | 0.8596 | 0.9783 | 15.74 | 38.57 | 71.49 | 34.02 | 196.43 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window7_224_22kto1k_infer.tar) | +| SwinTransformer_large_patch4_window12_384[1] | 0.8719 | 0.9823 | 32.61 | 116.59 | 223.23 | 99.97 | 196.43 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window12_384_22kto1k_infer.tar) | [1]：It is pre-trained based on the ImageNet22k dataset, and then transferred and learned from the ImageNet1k dataset. -## 17. LeViT series +## 17. LeViT series [[33](#ref33)] The accuracy and speed indicators of LeViT series models are shown in the following table. For more introduction, please refer to: [LeViT series model documents](../models/LeViT_en.md). | Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(M) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | | ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | -| LeViT_128S | 0.7598 | 0.9269 | | | | 281 | 7.42 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/eViT_128S_infer.tar) | +| LeViT_128S | 0.7598 | 0.9269 | | | | 281 | 7.42 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_128S_infer.tar) | | LeViT_128 | 0.7810 | 0.9371 | | | | 365 | 8.87 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_128_infer.tar) | | LeViT_192 | 0.7934 | 0.9446 | | | | 597 | 10.61 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_192_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_192_infer.tar) | | LeViT_256 | 0.8085 | 0.9497 | | | | 1049 | 18.45 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_256_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_256_infer.tar) | @@ -431,7 +429,7 @@ The accuracy and speed indicators of LeViT series models are shown in the follow -## 18. Twins series +## 18. Twins series [[34](#ref34)] The accuracy and speed indicators of Twins series models are shown in the following table. For more introduction, please refer to: [Twins series model documents](../models/Twins_en.md). @@ -448,7 +446,7 @@ The accuracy and speed indicators of Twins series models are shown in the follow -## 19. HarDNet series +## 19. HarDNet series [[37](#ref37)] The accuracy and speed indicators of HarDNet series models are shown in the following table. For more introduction, please refer to: [HarDNet series model documents](../models/HarDNet_en.md). @@ -461,7 +459,7 @@ The accuracy and speed indicators of HarDNet series models are shown in the foll -## 20. DLA series +## 20. DLA series [[38](#ref38)] The accuracy and speed indicators of DLA series models are shown in the following table. For more introduction, please refer to: [DLA series model documents](../models/DLA_en.md). @@ -479,7 +477,7 @@ The accuracy and speed indicators of DLA series models are shown in the followin -## 21. RedNet series +## 21. RedNet series [[39](#ref39)] The accuracy and speed indicators of RedNet series models are shown in the following table. For more introduction, please refer to: [RedNet series model documents](../models/RedNet_en.md). @@ -493,7 +491,7 @@ The accuracy and speed indicators of RedNet series models are shown in the follo -## 22. TNT series +## 22. TNT series [[35](#ref35)] The accuracy and speed indicators of TNT series models are shown in the following table. For more introduction, please refer to: [TNT series model documents](../models/TNT_en.md). @@ -505,9 +503,53 @@ The accuracy and speed indicators of TNT series models are shown in the followin -## 23. Other models +## 23. CSWinTransformer series [[40](#ref40)] + +The accuracy and speed indicators of CSWinTransformer series models are shown in the following table. For more introduction, please refer to: [CSWinTransformer series model documents](../models/CSWinTransformer_en.md)。 + +| Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | +| ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| CSWinTransformer_tiny_224 | 0.8281 | 0.9628 | - | - | - | 4.1 | 22 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_tiny_224_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_tiny_224_infer.tar) | +| CSWinTransformer_small_224 | 0.8358 | 0.9658 | - | - | - | 6.4 | 35 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_small_224_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_small_224_infer.tar) | +| CSWinTransformer_base_224 | 0.8420 | 0.9692 | - | - | - | 14.3 | 77 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_224_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_base_224_infer.tar) | +| CSWinTransformer_large_224 | 0.8643 | 0.9799 | - | - | - | 32.2 | 173.3 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_224_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_large_224_infer.tar) | +| CSWinTransformer_base_384 | 0.8550 | 0.9749 | - | - |- | 42.2 | 77 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_384_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_base_384_infer.tar) | +| CSWinTransformer_large_384 | 0.8748 | 0.9833 | - | - | - | 94.7 | 173.3 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_384_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_large_384_infer.tar) | + + + + +## 24. PVTV2 series [[41](#ref41)] + +The accuracy and speed indicators of PVTV2 series models are shown in the following table. For more introduction, please refer to: [PVTV2 series model documents](../models/PVTV2_en.md)。 + +| Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | +| ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| PVT_V2_B0 | 0.705 | 0.902 | - | - | - | 0.53 | 3.7 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B0_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B0_infer.tar) | +| PVT_V2_B1 | 0.787 | 0.945 | - | - | - | 2.0 | 14.0 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B1_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B1_infer.tar) | +| PVT_V2_B2 | 0.821 | 0.960 | - | - | - | 3.9 | 25.4 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B2_infer.tar) | +| PVT_V2_B2_Linear | 0.821 | 0.961 | - | - | - | 3.8 | 22.6 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2_Linear_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B2_Linear_infer.tar) | +| PVT_V2_B3 | 0.831 | 0.965 | - | - |- | 6.7 | 45.2 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B3_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B3_infer.tar) | +| PVT_V2_B4 | 0.836 | 0.967 | - | - | - | 9.8 | 62.6 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B4_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B4_infer.tar) | +| PVT_V2_B5 | 0.837 | 0.966 | - | - | - | 11.4 | 82.0 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B5_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B5_infer.tar) | + + -The accuracy and speed indicators of AlexNet, SqueezeNet series, VGG series, DarkNet53 and other models are shown in the following table. For more information, please refer to: [Other model documents](../models/Others_en.md). +## 25. MobileViT series [[42](#ref42)] + +The accuracy and speed indicators of MobileViT series models are shown in the following table. For more introduction, please refer to:[MobileViT series model documents](../models/MobileViT_en.md) + +| Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(M) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | +| ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| MobileViT_XXS | 0.6867 | 0.8878 | - | - | - | 1849.35 | 5.59 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XXS_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileViT_XXS_infer.tar) | +| MobileViT_XS | 0.7454 | 0.9227 | - | - | - | 930.75 | 2.33 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XS_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileViT_XS_infer.tar) | +| MobileViT_S | 0.7814 | 0.9413 | - | - | - | 337.24 | 1.28 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_S_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileViT_S_infer.tar) | + + + +## 26. Other models + +The accuracy and speed indicators of AlexNet [[18](#ref18)], SqueezeNet series [[19](#ref19)], VGG series [[20](#ref20)], DarkNet53 [[21](#ref21)] and other models are shown in the following table. For more information, please refer to: [Other model documents](../models/Others_en.md). | Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | Pretrained Model Download Address | Inference Model Download Address | |------------------------|-----------|-----------|------------------|------------------|----------|-----------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------| @@ -519,3 +561,93 @@ The accuracy and speed indicators of AlexNet, SqueezeNet series, VGG series, Dar | VGG16 | 0.720 | 0.907 | 2.48 | 6.79 | 12.33 | 15.470 | 138.35 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG16_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/VGG16_infer.tar) | | VGG19 | 0.726 | 0.909 | 2.93 | 8.28 | 15.21 | 19.63 | 143.66 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG19_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/VGG19_infer.tar) | | DarkNet53 | 0.780 | 0.941 | 2.79 | 6.42 | 10.89 | 9.31 | 41.65 | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DarkNet53_pretrained.pdparams) | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DarkNet53_infer.tar) | + + + +## Reference + +[1] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778. + +[2] He T, Zhang Z, Zhang H, et al. Bag of tricks for image classification with convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2019: 558-567. + +[3] Howard A, Sandler M, Chu G, et al. Searching for mobilenetv3[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1314-1324. + +[4] Sandler M, Howard A, Zhu M, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 4510-4520. + +[5] Howard A G, Zhu M, Chen B, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[J]. arXiv preprint arXiv:1704.04861, 2017. + +[6] Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 116-131. + +[7] Xie S, Girshick R, Dollár P, et al. Aggregated residual transformations for deep neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1492-1500. + +[8] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141. + +[9] Gao S, Cheng M M, Zhao K, et al. Res2net: A new multi-scale backbone architecture[J]. IEEE transactions on pattern analysis and machine intelligence, 2019. + +[10] Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 1-9. + +[11] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, inception-resnet and the impact of residual connections on learning[C]//Thirty-first AAAI conference on artificial intelligence. 2017. + +[12] Chollet F. Xception: Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1251-1258. + +[13] Wang J, Sun K, Cheng T, et al. Deep high-resolution representation learning for visual recognition[J]. arXiv preprint arXiv:1908.07919, 2019. + +[14] Chen Y, Li J, Xiao H, et al. Dual path networks[C]//Advances in neural information processing systems. 2017: 4467-4475. + +[15] Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 4700-4708. + +[16] Tan M, Le Q V. Efficientnet: Rethinking model scaling for convolutional neural networks[J]. arXiv preprint arXiv:1905.11946, 2019. + +[17] Mahajan D, Girshick R, Ramanathan V, et al. Exploring the limits of weakly supervised pretraining[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 181-196. + +[18] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]//Advances in neural information processing systems. 2012: 1097-1105. + +[19] Iandola F N, Han S, Moskewicz M W, et al. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size[J]. arXiv preprint arXiv:1602.07360, 2016. + +[20] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014. + +[21] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788. + +[22] Ding X, Guo Y, Ding G, et al. Acnet: Strengthening the kernel skeletons for powerful cnn via asymmetric convolution blocks[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1911-1920. + +[23] Han K, Wang Y, Tian Q, et al. GhostNet: More features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 1580-1589. + +[24] Zhang H, Wu C, Zhang Z, et al. Resnest: Split-attention networks[J]. arXiv preprint arXiv:2004.08955, 2020. + +[25] Radosavovic I, Kosaraju R P, Girshick R, et al. Designing network design spaces[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 10428-10436. + +[26] C.Szegedy, V.Vanhoucke, S.Ioffe, J.Shlens, and Z.Wojna. Rethinking the inception architecture for computer vision. arXiv preprint arXiv:1512.00567, 2015. + +[27] Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin and Baining Guo. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. + +[28]Cheng Cui, Tingquan Gao, Shengyu Wei, Yuning Du, Ruoyu Guo, Shuilong Dong, Bin Lu, Ying Zhou, Xueying Lv, Qiwen Liu, Xiaoguang Hu, Dianhai Yu, Yanjun Ma. PP-LCNet: A Lightweight CPU Convolutional Neural Network. + +[29]Mingxing Tan, Quoc V. Le. MixConv: Mixed Depthwise Convolutional Kernels. + +[30]Dongyoon Han, Sangdoo Yun, Byeongho Heo, YoungJoon Yoo. Rethinking Channel Dimensions for Efficient Model Design. + +[31]Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby. AN IMAGE IS WORTH 16X16 WORDS: +TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE. + +[32]Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Herve Jegou. Training data-efficient image transformers & distillation through attention. + +[33]Benjamin Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Herve Jegou, Matthijs Douze. LeViT: a Vision Transformer in ConvNet’s Clothing for Faster Inference. + +[34]Xiangxiang Chu, Zhi Tian, Yuqing Wang, Bo Zhang, Haibing Ren, Xiaolin Wei, Huaxia Xia, Chunhua Shen. Twins: Revisiting the Design of Spatial Attention in Vision Transformers. + +[35]Kai Han, An Xiao, Enhua Wu, Jianyuan Guo, Chunjing Xu, Yunhe Wang. Transformer in Transformer. + +[36]Xiaohan Ding, Xiangyu Zhang, Ningning Ma, Jungong Han, Guiguang Ding, Jian Sun. RepVGG: Making VGG-style ConvNets Great Again. + +[37]Ping Chao, Chao-Yang Kao, Yu-Shan Ruan, Chien-Hsiang Huang, Youn-Long Lin. HarDNet: A Low Memory Traffic Network. + +[38]Fisher Yu, Dequan Wang, Evan Shelhamer, Trevor Darrell. Deep Layer Aggregation. + +[39]Duo Lim Jie Hu, Changhu Wang, Xiangtai Li, Qi She, Lei Zhu, Tong Zhang, Qifeng Chen. Involution: Inverting the Inherence of Convolution for Visual Recognition. + + +[40]Xiaoyi Dong, Jianmin Bao, Dongdong Chen, Weiming Zhang, Nenghai Yu, Lu Yuan, Dong Chen, Baining Guo. CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows. + +[41]Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao. PVTv2: Improved Baselines with Pyramid Vision Transformer. + +[42]Sachin Mehta, Mohammad Rastegari. MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer. diff --git a/docs/en/algorithm_introduction/image_classification_en.md b/docs/en/algorithm_introduction/image_classification_en.md index fa2319c105e50f467a7c9bb078d8d09516890892..450cbc791914cd474141ccea402ab4282a358a95 100644 --- a/docs/en/algorithm_introduction/image_classification_en.md +++ b/docs/en/algorithm_introduction/image_classification_en.md @@ -55,7 +55,7 @@ After the data is settled, the model often determines the upper limit of the fin ### 2.3 Train the Model -After preparing the data and model, you can start training the model and updating the parameters of the model. After many iterations, a trained model can finally be obtained for image classification tasks. The training process of image classification requires a lot of experience and involves the setting of many hyperparameters. PaddleClas provides a series of [training tuning methods](https://github.com/PaddlePaddle/PaddleClas/blob/develop/docs/en/models/Tricks_en.md), which can help you quickly obtain a high-precision model. +After preparing the data and model, you can start training the model and updating the parameters of the model. After many iterations, a trained model can finally be obtained for image classification tasks. The training process of image classification requires a lot of experience and involves the setting of many hyperparameters. PaddleClas provides a series of [training tuning methods](https://github.com/PaddlePaddle/PaddleClas/blob/develop/docs/en/models_training/train_strategy_en.md), which can help you quickly obtain a high-precision model. diff --git a/docs/en/faq_series/faq_2020_s1_en.md b/docs/en/faq_series/faq_2020_s1_en.md index de5a50c87233296f2a2543600ba1c7d8fb16671c..3b2bda758e0f60b37bffa3b4ee3a5d772ff921d4 100644 --- a/docs/en/faq_series/faq_2020_s1_en.md +++ b/docs/en/faq_series/faq_2020_s1_en.md @@ -108,7 +108,7 @@ PaddleClas strictly follows the resolution used by the authors of the paper. Sin **A**: -There are many ssld pre-training models available in PaddleClas, which obtain better pre-training weights by semi-supervised knowledge distillation, so that the accuracy can be improved by replacing the ssld pre-training models with higher accuracy in transfer tasks or downstream vision tasks without replacing the structure files. For example, in PaddleSeg, [HRNet](../models/HRNet_en.md) , with the weight of the ssld pre-training model, achieves much better accuracy than other same models in the industry; In PaddleDetection, [PP- YOLO](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/configs/ppyolo/README_cn.md)with ssld pre-training weights has further improvement in the already high baseline. The transfer of classification with ssld pre-training weights also yields impressive results, and the benefits of knowledge distillation for the transfer of classification task is detailed in [SSLD Distillation Strategy](../advanced_tutorials/knowledge_distillation_en.md) +There are many ssld pre-training models available in PaddleClas, which obtain better pre-training weights by semi-supervised knowledge distillation, so that the accuracy can be improved by replacing the ssld pre-training models with higher accuracy in transfer tasks or downstream vision tasks without replacing the structure files. For example, in PaddleSeg, [HRNet](../models/HRNet_en.md) , with the weight of the ssld pre-training model, achieves much better accuracy than other same models in the industry; In PaddleDetection, [PP- YOLO](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/configs/ppyolo/README_cn.md)with ssld pre-training weights has further improvement in the already high baseline. The transfer of classification with ssld pre-training weights also yields impressive results, and the benefits of knowledge distillation for the transfer of classification task is detailed in [SSLD Distillation Strategy](../advanced_tutorials/distillation/distillation_en.md) @@ -143,7 +143,7 @@ When adopting multiple models for inference, it is recommended to first export t **A**： -- You can adopt auto-mixed precision training, which can gain a significantly faster speed with almost zero precision loss. Take ResNet50 as an example, the configuration file of auto-mixed precision training in PaddleClas can be found at: [ResNet50_fp16.yml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_fp16.yaml). The main step is to add the following lines to the standard configuration file. +- You can adopt auto-mixed precision training, which can gain a significantly faster speed with almost zero precision loss. Take ResNet50 as an example, the configuration file of auto-mixed precision training in PaddleClas can be found at: [ResNet50_amp_O1.yml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml). The main step is to add the following lines to the standard configuration file. ``` # mixed precision training @@ -351,7 +351,7 @@ At this stage, it has become a common practice in the image recognition field to **A**: If the existing strategy cannot further improve the accuracy of the model, it means that the model has almost reached saturation with the existing dataset and strategy, and two methods are provided here. - Mining relevant data: Use the model trained on the existing dataset to make predictions on the relevant data, label the data with higher confidence and add it to the training set for further training. Repeat the steps above to further improve the accuracy of the model. -- Knowledge distillation: You can use a larger model to train a teacher model with higher accuracy on the dataset, and then adopt the teacher model to teach a Student model, where the Student model is the target model. PaddleClas provides Baidu's own SSLD knowledge distillation scheme, which can steadily improve by more than 3% even on such a challenging classification task as ImageNet-1k. For the chapter on SSLD knowledge distillation, please refer to [**SSLD Knowledge Distillation**](../advanced_tutorials/knowledge_distillation_en.md). +- Knowledge distillation: You can use a larger model to train a teacher model with higher accuracy on the dataset, and then adopt the teacher model to teach a Student model, where the Student model is the target model. PaddleClas provides Baidu's own SSLD knowledge distillation scheme, which can steadily improve by more than 3% even on such a challenging classification task as ImageNet-1k. For the chapter on SSLD knowledge distillation, please refer to [**SSLD Knowledge Distillation**](../advanced_tutorials/distillation/distillation_en.md). diff --git a/docs/en/faq_series/faq_2021_s2_en.md b/docs/en/faq_series/faq_2021_s2_en.md index c18da5b094c13cbe4fb07c28753ab375d79d0bb4..2bd95485552ea983e2661fd623684f17f6a5efba 100644 --- a/docs/en/faq_series/faq_2021_s2_en.md +++ b/docs/en/faq_series/faq_2021_s2_en.md @@ -248,7 +248,7 @@ PaddleClas saves/updates the following three types of models during training. #### Q2.4.2: How can recognition models be fine-tuned to train on the basis of pre-trained models? -**A**：The fine-tuning training of the recognition model is similar to that of the classification model. The recognition model can be loaded with a pre-trained model of the product, and the training process can be found in [recognition model training](../../models_training/recognition_en.md), and we will continue to refine the documentation. +**A**：The fine-tuning training of the recognition model is similar to that of the classification model. The recognition model can be loaded with a pre-trained model of the product, and the training process can be found in [recognition model training](../models_training/recognition_en.md), and we will continue to refine the documentation. #### Q2.4.3: Why does it fail to run all mini-batches in each epoch when training metric learning? @@ -353,4 +353,4 @@ pip install paddle2onnx - `InputSpec()` function is used to describe the signature information of the model input, including the `shape`, `type` and `name` of the input data (can be omitted). - The `paddle.onnx.export()` function needs to specify the model grouping object `net`, the save path of the exported model `save_path`, and the description of the model's input data `input_spec`. - Note that the `paddlepaddle` `2.0.0` or above should be adopted.See [paddle.onnx.export](https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/onnx/) for more details on the parameters of the `paddle.onnx.export()` function. + Note that the `paddlepaddle` `2.0.0` or above should be adopted.See [paddle.onnx.export](https://www.paddlepaddle.org.cn/documentation/docs/en/api/paddle/onnx/export_en.html) for more details on the parameters of the `paddle.onnx.export()` function. diff --git a/docs/en/faq_series/faq_selected_30_en.md b/docs/en/faq_series/faq_selected_30_en.md index 54f6ca18182b6bf40dc37932f0451c657e02b258..535195df01ababc07afdad97a487bd6bbc73fd86 100644 --- a/docs/en/faq_series/faq_selected_30_en.md +++ b/docs/en/faq_series/faq_selected_30_en.md @@ -27,8 +27,8 @@ > > -- Q: 怎样根据自己的任务选择合适的模型进行训练？How to choose the right training model? -- A: If you want to deploy on the server with a high requirement for accuracy but not model storage size or prediction speed, then it is recommended to use ResNet_vd, Res2Net_vd, DenseNet, Xception, etc., which are suitable for server-side models. If you want to deploy on the mobile side, then it is recommended to use MobileNetV3 and GhostNet. Meanwhile, we suggest you refer to the speed-accuracy metrics chart in [Model Library](../models/models_intro_en.md) when choosing models. +- Q: How to choose the right training model? +- A: If you want to deploy on the server with a high requirement for accuracy but not model storage size or prediction speed, then it is recommended to use ResNet_vd, Res2Net_vd, DenseNet, Xception, etc., which are suitable for server-side models. If you want to deploy on the mobile side, then it is recommended to use MobileNetV3 and GhostNet. Meanwhile, we suggest you refer to the speed-accuracy metrics chart in [Model Library](../algorithm_introduction/ImageNet_models_en.md) when choosing models. > > @@ -280,7 +280,7 @@ Loss: > > - Q: How to train with Automatic Mixed Precision (AMP) during training? -- A: You can refer to [ResNet50_fp16.yaml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_fp16.yaml). Specifically, if you want your configuration file to support automatic mixed precision during model training, you can add the following information to the file. +- A: You can refer to [ResNet50_amp_O1.yaml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml). Specifically, if you want your configuration file to support automatic mixed precision during model training, you can add the following information to the file. ``` # mixed precision training diff --git a/docs/en/inference_deployment/paddle_hub_serving_deploy_en.md b/docs/en/inference_deployment/paddle_hub_serving_deploy_en.md index 56e68faad571ee13fae3acd8465434d63b5eda0b..c89142911f12ffcb2622fb8b5912cd9c960e56c4 100644 --- a/docs/en/inference_deployment/paddle_hub_serving_deploy_en.md +++ b/docs/en/inference_deployment/paddle_hub_serving_deploy_en.md @@ -48,7 +48,7 @@ Before installing the service module, you need to prepare the inference model an **Notice**: * The model file path can be viewed and modified in `PaddleClas/deploy/hubserving/clas/params.py`. * It should be noted that the prefix of model structure file and model parameters file must be `inference`. -* More models provided by PaddleClas can be obtained from the [model library](../models/models_intro_en.md). You can also use models trained by yourself. +* More models provided by PaddleClas can be obtained from the [model library](../algorithm_introduction/ImageNet_models_en.md). You can also use models trained by yourself. ## 4. Install Service Module diff --git a/docs/en/inference_deployment/paddle_lite_deploy_en.md b/docs/en/inference_deployment/paddle_lite_deploy_en.md index 12d45dd512a022850f87a922d28e19f5affd56f4..8fe92ee18fa80c30522cf80a38403ffd6078c0d5 100644 --- a/docs/en/inference_deployment/paddle_lite_deploy_en.md +++ b/docs/en/inference_deployment/paddle_lite_deploy_en.md @@ -4,7 +4,7 @@ This tutorial will introduce how to use [Paddle-Lite](https://github.com/PaddleP Paddle-Lite is a lightweight inference engine for PaddlePaddle. It provides efficient inference capabilities for mobile phones and IoTs, and extensively integrates cross-platform hardware to provide lightweight deployment solutions for mobile-side deployment issues. -If you only want to test speed, please refer to [The tutorial of Paddle-Lite mobile-side benchmark test](../extension/paddle_mobile_inference_en.md). +If you only want to test speed, please refer to [The tutorial of Paddle-Lite mobile-side benchmark test](../others/paddle_mobile_inference_en.md). --- @@ -18,6 +18,7 @@ If you only want to test speed, please refer to [The tutorial of Paddle-Lite mob - [2.1.1 [RECOMMEND] Use pip to install Paddle-Lite and optimize model](#2.1.1) - [2.1.2 Compile Paddle-Lite to generate opt tool](#2.1.2) - [2.1.3 Demo of get the optimized model](#2.1.3) + - [2.1.4 Compile to get the executable file clas_system](#2.1.4) - [2.2 Run optimized model on Phone](#2.2) - [3. FAQ](#3) @@ -40,20 +41,20 @@ For the detailed compilation directions of different development environments, p |Platform|Inference Library Download Link| |-|-| -|Android|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv7.gcc.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv8.gcc.c++_static.with_extra.with_cv.tar.gz)| -|iOS|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| +|Android|[arm7](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.android.armv7.clang.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv.tar.gz) | +|iOS|[arm7](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| **NOTE**: 1. If you download the inference library from [Paddle-Lite official document](https://paddle-lite.readthedocs.io/zh/latest/quick_start/release_lib.html#android-toolchain-gcc), please choose `with_extra=ON` , `with_cv=ON` . -2. It is recommended to build inference library using [Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) develop branch if you want to deploy the [quantitative](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/deploy/slim/quantization/README_en.md) model to mobile phones. Please refer to the [link](https://paddle-lite.readthedocs.io/zh/latest/user_guides/Compile/Android.html#id2) for more detailed information about compiling. +2. It is recommended to build inference library using [Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) develop branch if you want to deploy the [quantitative](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/deploy/slim/quantization/README_en.md) model to mobile phones. Please refer to the [link](https://paddle-lite.readthedocs.io/) for more detailed information about compiling. The structure of the inference library is as follows: ``` -inference_lite_lib.android.armv8/ +inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/ |-- cxx C++ inference library and header files | |-- include C++ header files | | |-- paddle_api.h @@ -148,6 +149,23 @@ paddle_lite_opt --model_file=./MobileNetV3_large_x1_0_infer/inference.pdmodel -- ``` When the above code command is completed, there will be ``MobileNetV3_large_x1_0.nb` in the current directory, which is the converted model file. + + +#### 2.1.4 Compile to get the executable file clas_system + +```shell +# Clone the Autolog repository to get automation logs +cd PaddleClas_root_path +cd deploy/lite/ +git clone https://github.com/LDOUBLEV/AutoLog.git +``` + +```shell +# Compile +make -j +``` + +After executing the `make` command, the `clas_system` executable file is generated in the current directory, which is used for Lite prediction. ## 2.2 Run optimized model on Phone @@ -172,7 +190,7 @@ When the above code command is completed, there will be ``MobileNetV3_large_x1_0 * Install ADB for windows If install ADB fo Windows, you need to download from Google's Android platform: [Download Link](https://developer.android.com/studio). - First, make sure the phone is connected to the computer, turn on the `USB debugging` option of the phone, and select the `file transfer` mode. Verify whether ADB is installed successfully as follows: +3. First, make sure the phone is connected to the computer, turn on the `USB debugging` option of the phone, and select the `file transfer` mode. Verify whether ADB is installed successfully as follows: ```shell $ adb devices @@ -183,42 +201,22 @@ When the above code command is completed, there will be ``MobileNetV3_large_x1_0 If there is `device` output like the above, it means the installation was successful. -4. Prepare optimized model, inference library files, test image and dictionary file used. +4. Push the optimized model, prediction library file, test image and class map file to the phone. ```shell -cd PaddleClas_root_path -cd deploy/lite/ - -# prepare.sh will put the inference library files, the test image and the dictionary files in demo/cxx/clas -sh prepare.sh /{lite inference library path}/inference_lite_lib.android.armv8 - -# enter the working directory of lite demo -cd /{lite inference library path}/inference_lite_lib.android.armv8/ -cd demo/cxx/clas/ - -# copy the C++ inference dynamic library file （ie. .so) to the debug folder -cp ../../../cxx/lib/libpaddle_light_api_shared.so ./debug/ +```shell +adb shell mkdir -p /data/local/tmp/arm_cpu/ +adb push clas_system /data/local/tmp/arm_cpu/ +adb shell chmod +x /data/local/tmp/arm_cpu//clas_system +adb push inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/cxx/lib/libpaddle_light_api_shared.so /data/local/tmp/arm_cpu/ +adb push MobileNetV3_large_x1_0.nb /data/local/tmp/arm_cpu/ +adb push config.txt /data/local/tmp/arm_cpu/ +adb push ../../ppcls/utils/imagenet1k_label_list.txt /data/local/tmp/arm_cpu/ +adb push imgs/tabby_cat.jpg /data/local/tmp/arm_cpu/ ``` -The `prepare.sh` take `PaddleClas/deploy/lite/imgs/tabby_cat.jpg` as the test image, and copy it to the `demo/cxx/clas/debug/` directory. - You should put the model that optimized by `paddle_lite_opt` under the `demo/cxx/clas/debug/` directory. In this example, use `MobileNetV3_large_x1_0.nb` model file generated in [2.1.3](#2.1.3). -The structure of the clas demo is as follows after the above command is completed: - -``` -demo/cxx/clas/ -|-- debug/ -| |--MobileNetV3_large_x1_0.nb class model -| |--tabby_cat.jpg test image -| |--imagenet1k_label_list.txt dictionary file -| |--libpaddle_light_api_shared.so C++ .so file -| |--config.txt config file -|-- config.txt config file -|-- image_classfication.cpp source code -|-- Makefile compile file -``` - **NOTE**: * `Imagenet1k_label_list.txt` is the category mapping file of the `ImageNet1k` dataset. If use a custom category, you need to replace the category mapping file. @@ -229,33 +227,22 @@ clas_model_file ./MobileNetV3_large_x1_0.nb # path of model file label_path ./imagenet1k_label_list.txt # path of category mapping file resize_short_size 256 # the short side length after resize crop_size 224 # side length used for inference after cropping - visualize 0 # whether to visualize. If you set it to 1, an image file named 'clas_result.png' will be generated in the current directory. +num_threads 1 # The number of threads, the default is 1 +precision FP32 # Precision type, you can choose FP32 or INT8, the default is FP32 +runtime_device arm_cpu # Device type, the default is arm_cpu +enable_benchmark 0 # Whether to enable benchmark, the default is 0 +tipc_benchmark 0 # Whether to enable tipc_benchmark, the default is 0 ``` 5. Run Model on Phone -```shell -# run compile to get the executable file 'clas_system' -make -j - -# move the compiled executable file to the debug folder -mv clas_system ./debug/ - -# push the debug folder to Phone -adb push debug /data/local/tmp/ +Execute the following command to complete the prediction on the mobile phone. -adb shell -cd /data/local/tmp/debug -export LD_LIBRARY_PATH=/data/local/tmp/debug:$LD_LIBRARY_PATH - -# the usage of clas_system is as follows: -# ./clas_system "path of config file" "path of test image" -./clas_system ./config.txt ./tabby_cat.jpg +```shell +adb shell 'export LD_LIBRARY_PATH=/data/local/tmp/arm_cpu/; /data/local/tmp/arm_cpu/clas_system /data/local/tmp/arm_cpu/config.txt /data/local/tmp/arm_cpu/tabby_cat.jpg' ``` -**NOTE**: If you make changes to the code, you need to recompile and repush the `debug ` folder to the phone. - The result is as follows: ![](../../images/inference_deployment/lite_demo_result.png) diff --git a/docs/en/models/CSWinTransformer_en.md b/docs/en/models/CSWinTransformer_en.md new file mode 100644 index 0000000000000000000000000000000000000000..2c9ef855a28a4db641ffa09c8f0f2d3d91571966 --- /dev/null +++ b/docs/en/models/CSWinTransformer_en.md @@ -0,0 +1,24 @@ +# CSWinTransformer +--- +## Catalogue + +* [1. Overview](#1) +* [2. Accuracy, FLOPs and Parameters](#2) + + + +## 1. Overview +CSWinTransformer is a new visual Transformer network that can be used as a general backbone network in the field of computer vision. CSWinTransformer proposes to do self-attention through a cross-shaped window, which not only has a very high computational efficiency, but also can obtain a global receptive field through two-layer calculation. CSWinTransformer also proposed a new encoding method: LePE, which further improved the accuracy of the model. [Paper](https://arxiv.org/abs/2107.00652) + + + +## 2. Accuracy, FLOPs and Parameters + +| Models | Top1 | Top5 | Reference
top1 | Reference
top5 | FLOPs
(G) | Params
(M) | +|:--:|:--:|:--:|:--:|:--:|:--:|:--:| +| CSWinTransformer_tiny_224 | 0.8281 | 0.9628 | 0.828 | - | 4.1 | 22 | +| CSWinTransformer_small_224 | 0.8358 | 0.9658 | 0.836 | - | 6.4 | 35 | +| CSWinTransformer_base_224 | 0.8420 | 0.9692 | 0.842 | - | 14.3 | 77 | +| CSWinTransformer_large_224 | 0.8643 | 0.9799 | 0.865 | - | 32.2 | 173.3 | +| CSWinTransformer_base_384 | 0.8550 | 0.9749 | 0.855 | - | 42.2 | 77 | +| CSWinTransformer_large_384 | 0.8748 | 0.9833 | 0.875 | - | 94.7 | 173.3 | diff --git a/docs/en/models/MobileViT_en.md b/docs/en/models/MobileViT_en.md new file mode 100644 index 0000000000000000000000000000000000000000..96b5e8260e51cf062b7da74b449140eb1bc68dd0 --- /dev/null +++ b/docs/en/models/MobileViT_en.md @@ -0,0 +1,23 @@ +# MobileviT +--- +## Catalogue + +* [1. Overview](#1) +* [2. Accuracy, FLOPs and Parameters](#2) + + + +## 1. Overview + +MobileViT is a lightweight visual Transformer network that can be used as a general backbone network in the field of computer vision. MobileViT combines the advantages of CNN and Transformer, which can better deal with global features and local features, and better solve the problem of lack of inductive bias in Transformer models. +, and finally, under the same amount of parameters, compared with other SOTA models, the tasks of image classification, object detection, and semantic segmentation have been greatly improved. [Paper](https://arxiv.org/pdf/2110.02178.pdf) + + + +## 2. Accuracy, FLOPs and Parameters + +| Models | Top1 | Top5 | Reference
top1 | Reference
top5 | FLOPs
(M) | Params
(M) | +|:--:|:--:|:--:|:--:|:--:|:--:|:--:| +| MobileViT_XXS | 0.6867 | 0.8878 | 0.690 | - | 1849.35 | 5.59 | +| MobileViT_XS | 0.7454 | 0.9227 | 0.747 | - | 930.75 | 2.33 | +| MobileViT_S | 0.7814 | 0.9413 | 0.783 | - | 337.24 | 1.28 | diff --git a/docs/en/models/PP-LCNet_en.md b/docs/en/models/PP-LCNet_en.md index 9eac6522ab9ce0961bf6b2272d90bd50866ea422..12d43c9e159d615d3c6397f7af447460e5bf1b7e 100644 --- a/docs/en/models/PP-LCNet_en.md +++ b/docs/en/models/PP-LCNet_en.md @@ -107,7 +107,7 @@ For image classification, ImageNet dataset is adopted. Compared with the current | PPLCNet_x1_0_ssld | 3.0 | 161 | 74.39 | 92.09 | 2.46 | | PPLCNet_x2_5_ssld | 9.0 | 906 | 80.82 | 95.33 | 5.39 | -where `_ssld` represents the model after using `SSLD distillation`. For details about `SSLD distillation`, see [SSLD distillation](../advanced_tutorials/knowledge_distillation_en.md). +where `_ssld` represents the model after using `SSLD distillation`. For details about `SSLD distillation`, see [SSLD distillation](../advanced_tutorials/distillation/distillation_en.md). Performance comparison with other lightweight networks: @@ -190,7 +190,7 @@ Rather than holding on to perfect FLOPs and Params as academics do, PP-LCNet foc Reference to cite when you use PP-LCNet in a paper: ``` @misc{cui2021pplcnet, - title={PP-LCNet: A Lightweight CPU Convolutional Neural Network}, + title={PP-LCNet: A Lightweight CPU Convolutional Neural Network}, author={Cheng Cui and Tingquan Gao and Shengyu Wei and Yuning Du and Ruoyu Guo and Shuilong Dong and Bin Lu and Ying Zhou and Xueying Lv and Qiwen Liu and Xiaoguang Hu and Dianhai Yu and Yanjun Ma}, year={2021}, eprint={2109.15099}, diff --git a/docs/en/models/models_intro_en.md b/docs/en/models/models_intro_en.md deleted file mode 100644 index 8d35459b6bd7e4e4c51d8b47bba52758980d89d7..0000000000000000000000000000000000000000 --- a/docs/en/models/models_intro_en.md +++ /dev/null @@ -1,305 +0,0 @@ -# Model Library Overview - -## Overview - -Based on the ImageNet1k classification dataset, the 29 classification network structures supported by PaddleClas and the corresponding 134 image classification pretrained models are shown below. Training trick, a brief introduction to each series of network structures, and performance evaluation will be shown in the corresponding chapters. - -## Evaluation environment -* Arm evaluation environment is based on Snapdragon 855 (SD855). -* The GPU evaluation environment is based on V100 and TensorRT, and the evaluation script is as follows. - -```shell -#!/usr/bin/env bash - -export PYTHONPATH=$PWD:$PYTHONPATH - -python tools/infer/predict.py \ - --model_file='pretrained/infer/model' \ - --params_file='pretrained/infer/params' \ - --enable_benchmark=True \ - --model_name=ResNet50_vd \ - --use_tensorrt=True \ - --use_fp16=False \ - --batch_size=1 -``` - - -![](../../images/models/V100_benchmark/v100.fp32.bs1.main_fps_top1_s.png) - - -![](../../images/models/mobile_arm_top1.png) - - -![](../../images/models/V100_benchmark/v100.fp32.bs1.visiontransformer.png) -
- -> If you think this document is helpful to you, welcome to give a star to our project:[https://github.com/PaddlePaddle/PaddleClas](https://github.com/PaddlePaddle/PaddleClas) - - -## Pretrained model list and download address -- ResNet and ResNet_vd series - - ResNet series[[1](#ref1)]([paper link](http://openaccess.thecvf.com/content_cvpr_2016/html/He_Deep_Residual_Learning_CVPR_2016_paper.html)) - - [ResNet18](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet18_pretrained.pdparams) - - [ResNet34](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet34_pretrained.pdparams) - - [ResNet50](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_pretrained.pdparams) - - [ResNet101](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet101_pretrained.pdparams) - - [ResNet152](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet152_pretrained.pdparams) - - ResNet_vc、ResNet_vd series[[2](#ref2)]([paper link](https://arxiv.org/abs/1812.01187)) - - [ResNet50_vc](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vc_pretrained.pdparams) - - [ResNet18_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet18_vd_pretrained.pdparams) - - [ResNet34_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet34_vd_pretrained.pdparams) - - [ResNet34_vd_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet34_vd_ssld_pretrained.pdparams) - - [ResNet50_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_pretrained.pdparams) - - [ResNet50_vd_v2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_v2_pretrained.pdparams) - - [ResNet101_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet101_vd_pretrained.pdparams) - - [ResNet152_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet152_vd_pretrained.pdparams) - - [ResNet200_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet200_vd_pretrained.pdparams) - - [ResNet50_vd_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_ssld_pretrained.pdparams) - - [ResNet50_vd_ssld_v2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_ssld_v2_pretrained.pdparams) - - [Fix_ResNet50_vd_ssld_v2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Fix_ResNet50_vd_ssld_v2_pretrained.pdparams) - - [ResNet101_vd_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet101_vd_ssld_pretrained.pdparams) - - -- Mobile and Embedded Vision Applications Network series - - MobileNetV3 series[[3](#ref3)]([paper link](https://arxiv.org/abs/1905.02244)) - - [MobileNetV3_large_x0_35](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_35_pretrained.pdparams) - - [MobileNetV3_large_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams) - - [MobileNetV3_large_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_75_pretrained.pdparams) - - [MobileNetV3_large_x1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x1_0_pretrained.pdparams) - - [MobileNetV3_large_x1_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x1_25_pretrained.pdparams) - - [MobileNetV3_small_x0_35](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x0_35_pretrained.pdparams) - - [MobileNetV3_small_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x0_5_pretrained.pdparams) - - [MobileNetV3_small_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x0_75_pretrained.pdparams) - - [MobileNetV3_small_x1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x1_0_pretrained.pdparams) - - [MobileNetV3_small_x1_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x1_25_pretrained.pdparams) - - [MobileNetV3_large_x1_0_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x1_0_ssld_pretrained.pdparams) - - [MobileNetV3_large_x1_0_ssld_int8]()(coming soon) - - [MobileNetV3_small_x1_0_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x1_0_ssld_pretrained.pdparams) - - MobileNetV2 series[[4](#ref4)]([paper link](https://arxiv.org/abs/1801.04381)) - - [MobileNetV2_x0_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_25_pretrained.pdparams) - - [MobileNetV2_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_5_pretrained.pdparams) - - [MobileNetV2_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_75_pretrained.pdparams) - - [MobileNetV2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_pretrained.pdparams) - - [MobileNetV2_x1_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x1_5_pretrained.pdparams) - - [MobileNetV2_x2_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x2_0_pretrained.pdparams) - - [MobileNetV2_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_ssld_pretrained.pdparams) - - MobileNetV1 series[[5](#ref5)]([paper link](https://arxiv.org/abs/1704.04861)) - - [MobileNetV1_x0_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_x0_25_pretrained.pdparams) - - [MobileNetV1_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_x0_5_pretrained.pdparams) - - [MobileNetV1_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_x0_75_pretrained.pdparams) - - [MobileNetV1](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_pretrained.pdparams) - - [MobileNetV1_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_ssld_pretrained.pdparams) - - ShuffleNetV2 series[[6](#ref6)]([paper link](https://arxiv.org/abs/1807.11164)) - - [ShuffleNetV2_x0_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams) - - [ShuffleNetV2_x0_33](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams) - - [ShuffleNetV2_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams) - - [ShuffleNetV2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams) - - [ShuffleNetV2_x1_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams) - - [ShuffleNetV2_x2_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams) - - [ShuffleNetV2_swish](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams) - - GhostNet series[[23](#ref23)]([paper link](https://arxiv.org/pdf/1911.11907.pdf)) - - [GhostNet_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x0_5_pretrained.pdparams) - - [GhostNet_x1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_0_pretrained.pdparams) - - [GhostNet_x1_3](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_pretrained.pdparams) - - [GhostNet_x1_3_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_ssld_pretrained.pdparams) - - -- SEResNeXt and Res2Net series - - ResNeXt series[[7](#ref7)]([paper link](https://arxiv.org/abs/1611.05431)) - - [ResNeXt50_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_32x4d_pretrained.pdparams) - - [ResNeXt50_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_64x4d_pretrained.pdparams) - - [ResNeXt101_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x4d_pretrained.pdparams) - - [ResNeXt101_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_64x4d_pretrained.pdparams) - - [ResNeXt152_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_32x4d_pretrained.pdparams) - - [ResNeXt152_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_64x4d_pretrained.pdparams) - - ResNeXt_vd series - - [ResNeXt50_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_32x4d_pretrained.pdparams) - - [ResNeXt50_vd_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_64x4d_pretrained.pdparams) - - [ResNeXt101_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_32x4d_pretrained.pdparams) - - [ResNeXt101_vd_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_64x4d_pretrained.pdparams) - - [ResNeXt152_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_32x4d_pretrained.pdparams) - - [ResNeXt152_vd_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_64x4d_pretrained.pdparams) - - SE_ResNet_vd series[[8](#ref8)]([paper link](https://arxiv.org/abs/1709.01507)) - - [SE_ResNet18_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet18_vd_pretrained.pdparams) - - [SE_ResNet34_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet34_vd_pretrained.pdparams) - - [SE_ResNet50_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet50_vd_pretrained.pdparams) - - SE_ResNeXt series - - [SE_ResNeXt50_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_32x4d_pretrained.pdparams) - - [SE_ResNeXt101_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt101_32x4d_pretrained.pdparams) - - SE_ResNeXt_vd series - - [SE_ResNeXt50_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_vd_32x4d_pretrained.pdparams) - - [SENet154_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SENet154_vd_pretrained.pdparams) - - Res2Net series[[9](#ref9)]([paper link](https://arxiv.org/abs/1904.01169)) - - [Res2Net50_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_26w_4s_pretrained.pdparams) - - [Res2Net50_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_pretrained.pdparams) - - [Res2Net50_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_ssld_pretrained.pdparams) - - [Res2Net50_14w_8s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_14w_8s_pretrained.pdparams) - - [Res2Net101_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net101_vd_26w_4s_pretrained.pdparams) - - [Res2Net101_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net101_vd_26w_4s_ssld_pretrained.pdparams) - - [Res2Net200_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net200_vd_26w_4s_pretrained.pdparams) - - [Res2Net200_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net200_vd_26w_4s_ssld_pretrained.pdparams) - - -- Inception series - - GoogLeNet series[[10](#ref10)]([paper link](https://arxiv.org/pdf/1409.4842.pdf)) - - [GoogLeNet](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GoogLeNet_pretrained.pdparams) - - InceptionV3 series[[26](#ref26)]([paper link](https://arxiv.org/abs/1512.00567)) - - [InceptionV3](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV3_pretrained.pdparams) - - InceptionV4 series[[11](#ref11)]([paper link](https://arxiv.org/abs/1602.07261)) - - [InceptionV4](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV4_pretrained.pdparams) - - Xception series[[12](#ref12)]([paper link](http://openaccess.thecvf.com/content_cvpr_2017/html/Chollet_Xception_Deep_Learning_CVPR_2017_paper.html)) - - [Xception41](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_pretrained.pdparams) - - [Xception41_deeplab](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_deeplab_pretrained.pdparams) - - [Xception65](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_pretrained.pdparams) - - [Xception65_deeplab](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_deeplab_pretrained.pdparams) - - [Xception71](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception71_pretrained.pdparams) - - -- HRNet series - - HRNet series[[13](#ref13)]([paper link](https://arxiv.org/abs/1908.07919)) - - [HRNet_W18_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W18_C_pretrained.pdparams) - - [HRNet_W18_C_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W18_C_ssld_pretrained.pdparams) - - [HRNet_W30_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W30_C_pretrained.pdparams) - - [HRNet_W32_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W32_C_pretrained.pdparams) - - [HRNet_W40_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W40_C_pretrained.pdparams) - - [HRNet_W44_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W44_C_pretrained.pdparams) - - [HRNet_W48_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W48_C_pretrained.pdparams) - - [HRNet_W48_C_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W48_C_ssld_pretrained.pdparams) - - [HRNet_W64_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W64_C_pretrained.pdparams) - - [SE_HRNet_W64_C_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_HRNet_W64_C_ssld_pretrained.pdparams) - - -- DPN and DenseNet series - - DPN series[[14](#ref14)]([paper link](https://arxiv.org/abs/1707.01629)) - - [DPN68](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN68_pretrained.pdparams) - - [DPN92](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN92_pretrained.pdparams) - - [DPN98](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN98_pretrained.pdparams) - - [DPN107](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN107_pretrained.pdparams) - - [DPN131](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN131_pretrained.pdparams) - - DenseNet series[[15](#ref15)]([paper link](https://arxiv.org/abs/1608.06993)) - - [DenseNet121](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet121_pretrained.pdparams) - - [DenseNet161](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet161_pretrained.pdparams) - - [DenseNet169](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet169_pretrained.pdparams) - - [DenseNet201](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet201_pretrained.pdparams) - - [DenseNet264](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet264_pretrained.pdparams) - - -- EfficientNet and ResNeXt101_wsl series - - EfficientNet series[[16](#ref16)]([paper link](https://arxiv.org/abs/1905.11946)) - - [EfficientNetB0_small](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_small_pretrained.pdparams) - - [EfficientNetB0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_pretrained.pdparams) - - [EfficientNetB1](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB1_pretrained.pdparams) - - [EfficientNetB2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB2_pretrained.pdparams) - - [EfficientNetB3](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB3_pretrained.pdparams) - - [EfficientNetB4](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB4_pretrained.pdparams) - - [EfficientNetB5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB5_pretrained.pdparams) - - [EfficientNetB6](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB6_pretrained.pdparams) - - [EfficientNetB7](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB7_pretrained.pdparams) - - ResNeXt101_wsl series[[17](#ref17)]([paper link](https://arxiv.org/abs/1805.00932)) - - [ResNeXt101_32x8d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x8d_wsl_pretrained.pdparams) - - [ResNeXt101_32x16d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x16d_wsl_pretrained.pdparams) - - [ResNeXt101_32x32d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x32d_wsl_pretrained.pdparams) - - [ResNeXt101_32x48d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x48d_wsl_pretrained.pdparams) - - [Fix_ResNeXt101_32x48d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Fix_ResNeXt101_32x48d_wsl_pretrained.pdparams) - - - -- ResNeSt and RegNet series - - ResNeSt series[[24](#ref24)]([paper link](https://arxiv.org/abs/2004.08955)) - - [ResNeSt50_fast_1s1x64d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_fast_1s1x64d_pretrained.pdparams) - - [ResNeSt50](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_pretrained.pdparams) - - RegNet series[[25](#ref25)]([paper link](https://arxiv.org/abs/2003.13678)) - - [RegNetX_4GF](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetX_4GF_pretrained.pdparams) - - -- Transformer series - - Swin-transformer series[[27](#ref27)]([paper link](https://arxiv.org/pdf/2103.14030.pdf)) - - [SwinTransformer_tiny_patch4_window7_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_tiny_patch4_window7_224_pretrained.pdparams) - - [SwinTransformer_small_patch4_window7_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_small_patch4_window7_224_pretrained.pdparams) - - [SwinTransformer_base_patch4_window7_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_pretrained.pdparams) - - [SwinTransformer_base_patch4_window12_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_pretrained.pdparams) - - [SwinTransformer_base_patch4_window7_224_22k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_22k_pretrained.pdparams) - - [SwinTransformer_base_patch4_window7_224_22kto1k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_22kto1k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window12_384_22k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window12_384_22kto1k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window7_224_22k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window7_224_22kto1k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams) - - - -- Other models - - AlexNet series[[18](#ref18)]([paper link](https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf)) - - [AlexNet](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/AlexNet_pretrained.pdparams) - - SqueezeNet series[[19](#ref19)]([paper link](https://arxiv.org/abs/1602.07360)) - - [SqueezeNet1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_0_pretrained.pdparams) - - [SqueezeNet1_1](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_1_pretrained.pdparams) - - VGG series[[20](#ref20)]([paper link](https://arxiv.org/abs/1409.1556)) - - [VGG11](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG11_pretrained.pdparams) - - [VGG13](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG13_pretrained.pdparams) - - [VGG16](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG16_pretrained.pdparams) - - [VGG19](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG19_pretrained.pdparams) - - DarkNet series[[21](#ref21)]([paper link](https://arxiv.org/abs/1506.02640)) - - [DarkNet53](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DarkNet53_pretrained.pdparams) - -**Note**: The pretrained models of EfficientNetB1-B7 in the above models are transferred from [pytorch version of EfficientNet](https://github.com/lukemelas/EfficientNet-PyTorch), and the ResNeXt101_wsl series of pretrained models are transferred from [Official repo](https://github.com/facebookresearch/WSL-Images), the remaining pretrained models are obtained by training with the PaddlePaddle framework, and the corresponding training hyperparameters are given in configs. - -## References - - -[1] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778. - -[2] He T, Zhang Z, Zhang H, et al. Bag of tricks for image classification with convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2019: 558-567. - -[3] Howard A, Sandler M, Chu G, et al. Searching for mobilenetv3[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1314-1324. - -[4] Sandler M, Howard A, Zhu M, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 4510-4520. - -[5] Howard A G, Zhu M, Chen B, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[J]. arXiv preprint arXiv:1704.04861, 2017. - -[6] Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 116-131. - -[7] Xie S, Girshick R, Dollár P, et al. Aggregated residual transformations for deep neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1492-1500. - - -[8] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141. - - -[9] Gao S, Cheng M M, Zhao K, et al. Res2net: A new multi-scale backbone architecture[J]. IEEE transactions on pattern analysis and machine intelligence, 2019. - -[10] Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 1-9. - - -[11] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, inception-resnet and the impact of residual connections on learning[C]//Thirty-first AAAI conference on artificial intelligence. 2017. - -[12] Chollet F. Xception: Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1251-1258. - -[13] Wang J, Sun K, Cheng T, et al. Deep high-resolution representation learning for visual recognition[J]. arXiv preprint arXiv:1908.07919, 2019. - -[14] Chen Y, Li J, Xiao H, et al. Dual path networks[C]//Advances in neural information processing systems. 2017: 4467-4475. - -[15] Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 4700-4708. - - -[16] Tan M, Le Q V. Efficientnet: Rethinking model scaling for convolutional neural networks[J]. arXiv preprint arXiv:1905.11946, 2019. - -[17] Mahajan D, Girshick R, Ramanathan V, et al. Exploring the limits of weakly supervised pretraining[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 181-196. - -[18] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]//Advances in neural information processing systems. 2012: 1097-1105. - -[19] Iandola F N, Han S, Moskewicz M W, et al. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size[J]. arXiv preprint arXiv:1602.07360, 2016. - -[20] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014. - -[21] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788. - -[22] Ding X, Guo Y, Ding G, et al. Acnet: Strengthening the kernel skeletons for powerful cnn via asymmetric convolution blocks[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1911-1920. - -[23] Han K, Wang Y, Tian Q, et al. GhostNet: More features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 1580-1589. - -[24] Zhang H, Wu C, Zhang Z, et al. Resnest: Split-attention networks[J]. arXiv preprint arXiv:2004.08955, 2020. - -[25] Radosavovic I, Kosaraju R P, Girshick R, et al. Designing network design spaces[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 10428-10436. - -[26] C.Szegedy, V.Vanhoucke, S.Ioffe, J.Shlens, and Z.Wojna. Rethinking the inception architecture for computer vision. arXiv preprint arXiv:1512.00567, 2015. - -[27] Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin and Baining Guo. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. diff --git a/docs/en/models_training/classification_en.md b/docs/en/models_training/classification_en.md index 0a9bce2b04daeb7d030b90a929bc225815d6c3ca..d6a742bd4cdd9fb3eefec7545183177e5b723230 100644 --- a/docs/en/models_training/classification_en.md +++ b/docs/en/models_training/classification_en.md @@ -217,7 +217,7 @@ Some of the configurable evaluation parameters are described as follows: **Note：** When loading the model to be evaluated, you only need to specify the path of the model file stead of the suffix. PaddleClas will automatically add the `.pdparams` suffix, such as [3.1.3 Resume Training](#3.1.3). -When loading the model to be evaluated, you only need to specify the path of the model file stead of the suffix. PaddleClas will automatically add the `.pdparams` suffix, such as [3.1.3 Resume Training](https://github.com/PaddlePaddle/PaddleClas/blob/ develop/docs/zh_CN/models_training/classification.md#3.1.3). +When loading the model to be evaluated, you only need to specify the path of the model file stead of the suffix. PaddleClas will automatically add the `.pdparams` suffix, such as [3.1.3 Resume Training](../models_training/classification_en.md#3.1.3). diff --git a/docs/en/others/feature_visiualization_en.md b/docs/en/others/feature_visiualization_en.md index 1d20525c93fe830653b47413c97f1a0faebea0b2..438c99963dad359d34c2e8688b2dbefd92368d5b 100644 --- a/docs/en/others/feature_visiualization_en.md +++ b/docs/en/others/feature_visiualization_en.md @@ -27,7 +27,7 @@ The first step is to select the model to be studied, here we choose ResNet50. Co wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_pretrained.pdparams ``` -For other pre-training models and codes of network structure, please download [model library](../../../ppcls/arch/backbone/) and [pre-training models](../models/models_intro_en.md). +For other pre-training models and codes of network structure, please download [model library](../../../ppcls/arch/backbone/) and [pre-training models](../algorithm_introduction/ImageNet_models_en.md). diff --git a/docs/en/quick_start/quick_start_classification_professional_en.md b/docs/en/quick_start/quick_start_classification_professional_en.md index 10c2174063c9914ac2f97c03ce5bb789a116a9d9..840643aafe4e9d432331910fabbb42978105e645 100644 --- a/docs/en/quick_start/quick_start_classification_professional_en.md +++ b/docs/en/quick_start/quick_start_classification_professional_en.md @@ -75,9 +75,23 @@ python3 -m paddle.distributed.launch \ The highest accuracy of the validation set is around 0.415. -* ** Note** +Here, multiple GPUs are used for training. If only one GPU is used, please specify the GPU with the `CUDA_VISIBLE_DEVICES` setting, and specify the GPU with the `--gpus` setting, the same below. For example, to train with only GPU 0: + +```shell +export CUDA_VISIBLE_DEVICES=0 +python3 -m paddle.distributed.launch \ + --gpus="0" \ + tools/train.py \ + -c ./ppcls/configs/quick_start/professional/ResNet50_vd_CIFAR100.yaml \ + -o Global.output_dir="output_CIFAR" \ + -o Optimizer.lr.learning_rate=0.01 +``` + +* **Notice**: + +* The GPUs specified in `--gpus` can be a subset of the GPUs specified in `CUDA_VISIBLE_DEVICES`. +* Since the initial learning rate and batch-size need to maintain a linear relationship, when training is switched from 4 GPUs to 1 GPU, the total batch-size is reduced to 1/4 of the original, and the learning rate also needs to be reduced to 1/4 of the original, so changed the default learning rate from 0.04 to 0.01. - * If the number of GPU cards is not 4, the accuracy of the validation set may be different from 0.415. To maintain a comparable accuracy, you need to change the learning rate in the configuration file to `the current learning rate / 4 \* current card number`. The same below. diff --git a/docs/en/quick_start/quick_start_multilabel_classification_en.md b/docs/en/quick_start/quick_start_multilabel_classification_en.md index 39adbb7655b29f4bfe5f9545199d15fa6ac2876b..c2b1ddecd0d0d7d486858b682437364f6c4d674c 100644 --- a/docs/en/quick_start/quick_start_multilabel_classification_en.md +++ b/docs/en/quick_start/quick_start_multilabel_classification_en.md @@ -1,6 +1,6 @@ # Quick Start of Multi-label Classification -Experience the training, evaluation, and prediction of multi-label classification based on the [NUS-WIDE-SCENE](https://lms.comp.nus.edu.sg/wp-content/uploads/2019/research/nuswide/NUS-WIDE.html) dataset, which is a subset of the NUS-WIDE dataset. Please first install PaddlePaddle and PaddleClas, see [Paddle Installation](https://github.com/PaddlePaddle/PaddleClas/blob/develop/docs/zh_CN/installation) and [PaddleClas installation](https://github.com/PaddlePaddle/PaddleClas/blob/develop/docs/zh_CN/installation/install_ paddleclas.md) for more details. +Experience the training, evaluation, and prediction of multi-label classification based on the [NUS-WIDE-SCENE](https://lms.comp.nus.edu.sg/wp-content/uploads/2019/research/nuswide/NUS-WIDE.html) dataset, which is a subset of the NUS-WIDE dataset. Please first install PaddlePaddle and PaddleClas, see [Paddle Installation](../installation/install_paddle_en.md) and [PaddleClas installation](../installation/install_paddleclas_en.md) for more details. ## Catalogue diff --git a/docs/images/ISE_pipeline.png b/docs/images/ISE_pipeline.png new file mode 100644 index 0000000000000000000000000000000000000000..665a7e442a972a9d36f60c4e04a8a376bb270612 Binary files /dev/null and b/docs/images/ISE_pipeline.png differ diff --git a/docs/zh_CN/advanced_tutorials/gallery2fc.md b/docs/zh_CN/advanced_tutorials/gallery2fc.md new file mode 100644 index 0000000000000000000000000000000000000000..87d5caa81e87c9d4bd1c2d707d2b850994372488 --- /dev/null +++ b/docs/zh_CN/advanced_tutorials/gallery2fc.md @@ -0,0 +1,53 @@ +# 识别模型转分类模型 + +PaddleClas 提供了 `gallery2fc.py` 工具，帮助大家将识别模型转为分类模型。目前该工具仅支持转换量化后模型，因此建议使用 PaddleClas 提供的 `general_PPLCNet_x2_5_pretrained_v1.0_quant` 预训练模型，该模型为量化后的通用识别模型，backbone 为 PPLCNet_x2_5。 + +如需使用其他模型，关于量化的具体操作请参考文档 [模型量化](./model_prune_quantization.md)。 + +## 一、模型转换说明 + +### 1.1 准备底库数据、预训练模型 + +#### 1. 底库数据集 + +首先需要准备好底库数据，下面以 PaddleClas 提供的饮料数据集（drink_dataset_v1.0）为例进行说明，饮料数据集获取方法： + +```shell +cd PaddleClas/ +wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/data/drink_dataset_v1.0.tar +tar -xf drink_dataset_v1.0.tar +``` + +饮料数据集的底库图片路径为 `drink_dataset_v1.0/gallery/`，底库图片列表可在 `drink_dataset_v1.0/gallery/drink_label.txt` 中查看，关于底库数据格式说明，请参考文档[数据集格式说明](../data_preparation/recognition_dataset.md#1-数据集格式说明)。 + +#### 2. 预训练模型 + +在开始转换模型前，需要准备好预训练模型，下面以量化后的 `general_PPLCNet_x2_5` 模型为例，下载预训练模型： + +```shell +cd PaddleClas/pretrained/ +wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/pretrain/general_PPLCNet_x2_5_pretrained_v1.0_quant.pdparams +``` + +### 1.2 准备配置文件 + +在进行模型转换时，需要通过配置文件定义所需参数，本例中所用配置文件为 `ppcls/configs/GeneralRecognition/Gallery2FC_PPLCNet_x2_5.yaml`，对于配置文件字段的说明，如下所示： + +* Global: + * pretrained_model: 预训练模型路径，无需包含 `.pdparams` 后缀名； + * image_shape: 模型输入数据尺寸，无需包含 batch size 维度； + * save_inference_dir: 转换后模型的保存路径； +* Arch: 模型结构相关定义，可参考 [配置说明](../models_training/config_description.md#3-%E8%AF%86%E5%88%AB%E6%A8%A1%E5%9E%8B)； +* IndexProcess: 底库数据集相关定义 + * image_root: 底库数据集路径； + * data_file: 底库数据集列表文件路径； + +### 1.3 模型转换 + +在完成上述准备工作后，即可进行模型转换，命令如下所示： + +```python +python ppcls/utils/gallery2fc.py -c ppcls/configs/GeneralRecognition/Gallery2FC_PPLCNet_x2_5.yaml +``` + +在上述命令执行完成后，转换并导出的模型保存在目录 `./inference/general_PPLCNet_x2_5_quant/` 下。在推理部署时，需要注意的是，模型的输出结果通常有多个，应选取分类结果作为模型输出，需要注意区分。 diff --git a/docs/zh_CN/advanced_tutorials/shitu_deploy.md b/docs/zh_CN/advanced_tutorials/shitu_deploy.md index 064f68ee41671802ef65e83efee5565cf586a7c6..0693337000f36c100316e3bb12743b1396ea6a26 100644 --- a/docs/zh_CN/advanced_tutorials/shitu_deploy.md +++ b/docs/zh_CN/advanced_tutorials/shitu_deploy.md @@ -80,7 +80,7 @@ 因为要对模型进行训练，所以收集自己的数据集。数据准备及相应格式请参考：[特征提取文档](../image_recognition_pipeline/feature_extraction.md)中 `4.1数据准备`部分、[识别数据集说明](../data_preparation/recognition_dataset.md)。值得注意的是，此部分需要准备大量的数据，以保证识别模型效果。训练配置文件参考：[通用识别模型配置文件](../../../ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml)，训练方法参考：[识别模型训练](../models_training/recognition.md) - 数据增强：根据实际情况选择不同数据增强方法。如：实际应用中数据遮挡比较严重，建议添加`RandomErasing`增强方法。详见[数据增强文档](./DataAugmentation.md) -- 换不同的`backbone`，一般来说，越大的模型，特征提取能力更强。不同`backbone`详见[模型介绍](../models/models_intro.md) +- 换不同的`backbone`，一般来说，越大的模型，特征提取能力更强。不同`backbone`详见[模型介绍](../algorithm_introduction/ImageNet_models.md) - 选择不同的`Metric Learning`方法。不同的`Metric Learning`方法，对不同的数据集效果可能不太一样，建议尝试其他`Loss`,详见[Metric Learning](../algorithm_introduction/metric_learning.md) - 采用蒸馏方法，对小模型进行模型能力提升，详见[模型蒸馏](../algorithm_introduction/knowledge_distillation.md) - 增补数据集。针对错误样本，添加badcase数据 diff --git a/docs/zh_CN/algorithm_introduction/ISE_ReID.md b/docs/zh_CN/algorithm_introduction/ISE_ReID.md new file mode 100644 index 0000000000000000000000000000000000000000..48e52605535422cf92157d43a16cdf4e8d867645 --- /dev/null +++ b/docs/zh_CN/algorithm_introduction/ISE_ReID.md @@ -0,0 +1,62 @@ +# ISE +--- +## 目录 + +- [1. 介绍](#1) +- [2. 在Market1501和MSMT17上的结果](#2) +- [3. 测试](#3) +- [4. 引用](#4) + + +## 1. 介绍 + +ISE (Implicit Sample Extension)是一种简单、高效、有效的无监督行人再识别学习算法。ISE在聚类蔟边界周围生成样本，我们称之为支持样本。ISE的样本生成过程依赖于两个关键机制，即渐进线性插值策略（progressive linear interpolation）和标签保留的损失函数（label-preserving loss function）。ISE生成的支持样本提供了额外补充信息，可以很好地处理“子类和混合”的聚类错误。ISE在Market1501和MSMT17数据集上取得了优于其他无监督方法的性能。 + +> [**Implicit Sample Extension for Unsupervised Person Re-Identification**](https://arxiv.org/abs/2204.06892v1)
+> Xinyu Zhang, Dongdong Li, Zhigang Wang, Jian Wang, Errui Ding, Javen Qinfeng Shi, Zhaoxiang Zhang, Jingdong Wang
+> CVPR2022 + +![image](../../images/ISE_ReID/ISE_pipeline.png) + + + +## 2. 在Market1501和MSMT17上的结果 + +在Market1501和MSMT17上的主要结果。“PIL”表示渐进线性插值策略。“LP”表示标签保留的损失函数。 + +| 方法 | Market1501 | 下载链接 | MSMT17 | 下载链接 | +| --- | -- | -- | -- | - | +| Baseline | 82.5 (92.5) | - | 30.1 (58.6) | - | +| ISE (+PIL) | 83.9 (93.9) | - | 33.5 (63.9) | - | +| ISE (+LP) | 83.6 (92.7) | - | 31.4 (59.9) | - | +| ISE (Ours) (+PIL+LP) | **84.7 (94.0)** | [ISE_M](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ISE_M_model.pdparams) | **35.0 (64.7)** | [ISE_MS](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ISE_MS_model.pdparams) | + + + +## 3. 测试 +我们很快会提供训练代码，首先我们提供了测试代码和模型。 + +**测试：** 可简使用如下脚本进行模型评估。 + +``` +python tools/eval.py -c ./ppcls/configs/Person/ResNet50_UReID_infer.yaml +``` +**步骤：** +1. 首先下载模型，并放入：```./pd_model_trace/ISE/```。 +2. 改变```./ppcls/configs/Person/ResNet50_UReID_infer.yaml```中的数据集名称。 +3. 运行上述脚本。 + + + +## 4. 引用 + +如果ISE在您的研究中有启发，请考虑引用我们的论文: + +``` +@inproceedings{zhang2022Implicit, + title={Implicit Sample Extension for Unsupervised Person Re-Identification}, + author={Xinyu Zhang, Dongdong Li, Zhigang Wang, Jian Wang, Errui Ding, Javen Qinfeng Shi, Zhaoxiang Zhang, Jingdong Wang}, + booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, + year={2022} +} +``` diff --git a/docs/zh_CN/algorithm_introduction/ImageNet_models.md b/docs/zh_CN/algorithm_introduction/ImageNet_models.md index 0cbf58fccef6c196b3836366d39fe1c9e175d130..ee98de442a40fb7c37b2274b756a728f7dcfc5af 100644 --- a/docs/zh_CN/algorithm_introduction/ImageNet_models.md +++ b/docs/zh_CN/algorithm_introduction/ImageNet_models.md @@ -30,7 +30,11 @@ - [20. DLA 系列](#20) - [21. RedNet 系列](#21) - [22. TNT 系列](#22) -- [23. 其他模型](#23) +- [23. CSwinTransformer 系列](#23) +- [24. PVTV2 系列](#24) +- [25. MobileViT 系列](#25) +- [26. 其他模型](#26) +- [参考文献](#reference) @@ -65,7 +69,7 @@ | 模型 | Top-1 Acc | Reference
Top-1 Acc | Acc gain | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |---------------------|-----------|-----------|---------------|----------------|-----------|----------|-----------|-----------------------------------|-----------------------------------|-----------------------------------| -| ResNet34_vd_ssld | 0.797 | 0.760 | 0.037 | 2.00 | 3.28 | 5.84 | 3.93 | 21.84 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_ssld_pretrained.pdparams)   | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet34_vd_ssld.tar)   | +| ResNet34_vd_ssld | 0.797 | 0.760 | 0.037 | 2.00 | 3.28 | 5.84 | 3.93 | 21.84 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_ssld_pretrained.pdparams)   | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet34_vd_ssld_infer.tar)   | | ResNet50_vd_ssld | 0.830 | 0.792 | 0.039 | 2.60 | 4.86 | 7.63 | 4.35 | 25.63 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet50_vd_ssld_infer.tar) | | ResNet101_vd_ssld | 0.837 | 0.802 | 0.035 | 4.43 | 8.25 | 12.60 | 8.08 | 44.67 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_vd_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet101_vd_ssld_infer.tar) | | Res2Net50_vd_26w_4s_ssld | 0.831 | 0.798 | 0.033 | 3.59 | 6.35 | 9.50 | 4.28 | 25.76 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/Res2Net50_vd_26w_4s_ssld_infer.tar) | @@ -98,14 +102,11 @@ | PPLCNet_x1_0_ssld | 0.744 | 0.713 | 0.033 | 2.46 | 160.81 | 2.96 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x1_0_ssld_infer.tar) | | PPLCNet_x2_5_ssld | 0.808 | 0.766 | 0.042 | 5.39 | 906.49 | 9.04 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x2_5_ssld_infer.tar) | - - - * 注: `Reference Top-1 Acc` 表示 PaddleClas 基于 ImageNet1k 数据集训练得到的预训练模型精度。 -## 3. PP-LCNet 系列 +## 3. PP-LCNet 系列 [[28](#ref28)] PP-LCNet 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[PP-LCNet 系列模型文档](../models/PP-LCNet.md)。 @@ -122,7 +123,7 @@ PP-LCNet 系列模型的精度、速度指标如下表所示，更多关于该 -## 4. ResNet 系列 +## 4. ResNet 系列 [[1](#ref1)] ResNet 及其 Vd 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[ResNet 及其 Vd 系列模型文档](../models/ResNet_and_vd.md)。 @@ -146,7 +147,7 @@ ResNet 及其 Vd 系列模型的精度、速度指标如下表所示，更多关 -## 5. 移动端系列 +## 5. 移动端系列 [[3](#ref3)][[4](#ref4)][[5](#ref5)][[6](#ref6)][[23](#ref23)] 移动端系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[移动端系列模型文档](../models/Mobile.md)。 @@ -195,7 +196,7 @@ ResNet 及其 Vd 系列模型的精度、速度指标如下表所示，更多关 -## 6. SEResNeXt 与 Res2Net 系列 +## 6. SEResNeXt 与 Res2Net 系列 [[7](#ref7)][[8](#ref8)][[9](#ref9)] SEResNeXt 与 Res2Net 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[SEResNeXt 与 Res2Net 系列模型文档](../models/SEResNext_and_Res2Net.md)。 @@ -230,7 +231,7 @@ SEResNeXt 与 Res2Net 系列模型的精度、速度指标如下表所示，更 -## 7. DPN 与 DenseNet 系列 +## 7. DPN 与 DenseNet 系列 [[14](#ref14)][[15](#ref15)] DPN 与 DenseNet 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[DPN 与 DenseNet 系列模型文档](../models/DPN_DenseNet.md)。 @@ -248,15 +249,12 @@ DPN 与 DenseNet 系列模型的精度、速度指标如下表所示，更多关 | DPN107 | 0.8089 | 0.9532 | 19.46 | 35.62 | 50.22 | 18.38 | 87.13 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN107_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DPN107_infer.tar) | | DPN131 | 0.8070 | 0.9514 | 19.64 | 34.60 | 47.42 | 16.09 | 79.48 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN131_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DPN131_infer.tar) | - - -## 8. HRNet 系列 +## 8. HRNet 系列 [[13](#ref13)] HRNet 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[HRNet 系列模型文档](../models/HRNet.md)。 - | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |-------------|-----------|-----------|------------------|------------------|----------|-----------|--------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------| | HRNet_W18_C | 0.7692 | 0.9339 | 6.66 | 8.94 | 11.95 | 4.32 | 21.35 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W18_C_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/HRNet_W18_C_infer.tar) | @@ -272,7 +270,7 @@ HRNet 系列模型的精度、速度指标如下表所示，更多关于该系 -## 9. Inception 系列 +## 9. Inception 系列 [[10](#ref10)][[11](#ref11)][[12](#ref12)][[26](#ref26)] Inception 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[Inception 系列模型文档](../models/Inception.md)。 @@ -289,11 +287,10 @@ Inception 系列模型的精度、速度指标如下表所示，更多关于该 -## 10. EfficientNet 与 ResNeXt101_wsl 系列 +## 10. EfficientNet 与 ResNeXt101_wsl 系列 [[16](#ref16)][[17](#ref17)] EfficientNet 与 ResNeXt101_wsl 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[EfficientNet 与 ResNeXt101_wsl 系列模型文档](../models/EfficientNet_and_ResNeXt101_wsl.md)。 - | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |---------------------------|-----------|-----------|------------------|------------------|----------|-----------|----------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------| | ResNeXt101_
32x8d_wsl | 0.8255 | 0.9674 | 13.55 | 23.39 | 36.18 | 16.48 | 88.99 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x8d_wsl_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNeXt101_32x8d_wsl_infer.tar) | @@ -313,11 +310,10 @@ EfficientNet 与 ResNeXt101_wsl 系列模型的精度、速度指标如下表所 -## 11. ResNeSt 与 RegNet 系列 +## 11. ResNeSt 与 RegNet 系列 [[24](#ref24)][[25](#ref25)] ResNeSt 与 RegNet 系列模型的精度、速度指标如下表所示，更多关于该系列的模型介绍可以参考：[ResNeSt 与 RegNet 系列模型文档](../models/ResNeSt_RegNet.md)。 - | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |------------------------|-----------|-----------|------------------|------------------|----------|-----------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------| | ResNeSt50_
fast_1s1x64d | 0.8035 | 0.9528 | 2.73 | 5.33 | 8.24 | 4.36 | 26.27 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_fast_1s1x64d_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNeSt50_fast_1s1x64d_infer.tar) | @@ -326,11 +322,10 @@ ResNeSt 与 RegNet 系列模型的精度、速度指标如下表所示，更多 -## 12. ViT_and_DeiT 系列 +## 12. ViT_and_DeiT 系列 [[31](#ref31)][[32](#ref32)] ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模型的精度、速度指标如下表所示. 更多关于该系列模型的介绍可以参考： [ViT_and_DeiT 系列模型文档](../models/ViT_and_DeiT.md)。 - | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |------------------------|-----------|-----------|------------------|------------------|----------|------------------------|------------------------|------------------------|------------------------| | ViT_small_
patch16_224 | 0.7769 | 0.9342 | 3.71 | 9.05 | 16.72 | 9.41 | 48.60 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_small_patch16_224_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ViT_small_patch16_224_infer.tar) | @@ -341,8 +336,6 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 |ViT_large_
patch16_384| 0.8513 | 0.9736 | 39.51 | 152.46 | 304.06 | 174.70 | 304.12 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_384_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ViT_large_patch16_384_infer.tar) | |ViT_large_
patch32_384| 0.8153 | 0.9608 | 11.44 | 36.09 | 70.63 | 44.24 | 306.48 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch32_384_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ViT_large_patch32_384_infer.tar) | - - | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |------------------------|-----------|-----------|------------------|------------------|----------|------------------------|------------------------|------------------------|------------------------| | DeiT_tiny_
patch16_224 | 0.718 | 0.910 | 3.61 | 3.94 | 6.10 | 1.07 | 5.68 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_patch16_224_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DeiT_tiny_patch16_224_infer.tar) | @@ -356,11 +349,10 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 13. RepVGG 系列 +## 13. RepVGG 系列 [[36](#ref36)] 关于 RepVGG 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[RepVGG 系列模型文档](../models/RepVGG.md)。 - | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |------------------------|-----------|-----------|------------------|------------------|----------|-----------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------| | RepVGG_A0 | 0.7131 | 0.9016 | | | | 1.36 | 8.31 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A0_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/RepVGG_A0_infer.tar) | @@ -376,7 +368,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 14. MixNet 系列 +## 14. MixNet 系列 [[29](#ref29)] 关于 MixNet 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[MixNet 系列模型文档](../models/MixNet.md)。 @@ -388,7 +380,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 15. ReXNet 系列 +## 15. ReXNet 系列 [[30](#ref30)] 关于 ReXNet 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[ReXNet 系列模型文档](../models/ReXNet.md)。 @@ -402,7 +394,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 16. SwinTransformer 系列 +## 16. SwinTransformer 系列 [[27](#ref27)] 关于 SwinTransformer 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[SwinTransformer 系列模型文档](../models/SwinTransformer.md)。 @@ -414,20 +406,20 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 | SwinTransformer_base_patch4_window12_384 | 0.8439 | 0.9693 | 19.52 | 64.56 | 123.30 | 44.45 | 87.70 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_base_patch4_window12_384_infer.tar) | | SwinTransformer_base_patch4_window7_224[1] | 0.8487 | 0.9746 | 13.53 | 23.46 | 39.13 | 15.13 | 87.70 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_22kto1k_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_base_patch4_window7_224_infer.tar) | | SwinTransformer_base_patch4_window12_384[1] | 0.8642 | 0.9807 | 19.65 | 64.72 | 123.42 | 44.45 | 87.70 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_22kto1k_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_base_patch4_window12_384_infer.tar) | -| SwinTransformer_large_patch4_window7_224[1] | 0.8596 | 0.9783 | 15.74 | 38.57 | 71.49 | 34.02 | 196.43 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window7_224_infer.tar) | -| SwinTransformer_large_patch4_window12_384[1] | 0.8719 | 0.9823 | 32.61 | 116.59 | 223.23 | 99.97 | 196.43 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window12_384_infer.tar) | +| SwinTransformer_large_patch4_window7_224[1] | 0.8596 | 0.9783 | 15.74 | 38.57 | 71.49 | 34.02 | 196.43 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window7_224_22kto1k_infer.tar) | +| SwinTransformer_large_patch4_window12_384[1] | 0.8719 | 0.9823 | 32.61 | 116.59 | 223.23 | 99.97 | 196.43 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/SwinTransformer_large_patch4_window12_384_22kto1k_infer.tar) | [1]：基于 ImageNet22k 数据集预训练，然后在 ImageNet1k 数据集迁移学习得到。 -## 17. LeViT 系列 +## 17. LeViT 系列 [[33](#ref33)] 关于 LeViT 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[LeViT 系列模型文档](../models/LeViT.md)。 | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(M) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | | ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | -| LeViT_128S | 0.7598 | 0.9269 | | | | 281 | 7.42 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/eViT_128S_infer.tar) | +| LeViT_128S | 0.7598 | 0.9269 | | | | 281 | 7.42 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_128S_infer.tar) | | LeViT_128 | 0.7810 | 0.9371 | | | | 365 | 8.87 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_128_infer.tar) | | LeViT_192 | 0.7934 | 0.9446 | | | | 597 | 10.61 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_192_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_192_infer.tar) | | LeViT_256 | 0.8085 | 0.9497 | | | | 1049 | 18.45 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_256_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/LeViT_256_infer.tar) | @@ -437,7 +429,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 18. Twins 系列 +## 18. Twins 系列 [[34](#ref34)] 关于 Twins 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[Twins 系列模型文档](../models/Twins.md)。 @@ -454,7 +446,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 19. HarDNet 系列 +## 19. HarDNet 系列 [[37](#ref37)] 关于 HarDNet 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[HarDNet 系列模型文档](../models/HarDNet.md)。 @@ -467,7 +459,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 20. DLA 系列 +## 20. DLA 系列 [[38](#ref38)] 关于 DLA 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[DLA 系列模型文档](../models/DLA.md)。 @@ -485,7 +477,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 21. RedNet 系列 +## 21. RedNet 系列 [[39](#ref39)] 关于 RedNet 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[RedNet 系列模型文档](../models/RedNet.md)。 @@ -499,7 +491,7 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 22. TNT 系列 +## 22. TNT 系列 [[35](#ref35)] 关于 TNT 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[TNT 系列模型文档](../models/TNT.md)。 @@ -511,10 +503,54 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 -## 23. 其他模型 +## 23. CSWinTransformer 系列 [[40](#ref40)] + +关于 CSWinTransformer 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[CSWinTransformer 系列模型文档](../models/CSWinTransformer.md)。 + +| 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | +| ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| CSWinTransformer_tiny_224 | 0.8281 | 0.9628 | - | - | - | 4.1 | 22 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_tiny_224_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_tiny_224_infer.tar) | +| CSWinTransformer_small_224 | 0.8358 | 0.9658 | - | - | - | 6.4 | 35 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_small_224_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_small_224_infer.tar) | +| CSWinTransformer_base_224 | 0.8420 | 0.9692 | - | - | - | 14.3 | 77 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_224_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_base_224_infer.tar) | +| CSWinTransformer_large_224 | 0.8643 | 0.9799 | - | - | - | 32.2 | 173.3 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_224_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_large_224_infer.tar) | +| CSWinTransformer_base_384 | 0.8550 | 0.9749 | - | - |- | 42.2 | 77 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_384_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_base_384_infer.tar) | +| CSWinTransformer_large_384 | 0.8748 | 0.9833 | - | - | - | 94.7 | 173.3 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_384_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/CSWinTransformer_large_384_infer.tar) | + + + + +## 24. PVTV2 系列 [[41](#ref41)] -关于 AlexNet、SqueezeNet 系列、VGG 系列、DarkNet53 等模型的精度、速度指标如下表所示，更多介绍可以参考：[其他模型文档](../models/Others.md)。 +关于 PVTV2 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[PVTV2 系列模型文档](../models/PVTV2.md)。 +| 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | +| ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| PVT_V2_B0 | 0.705 | 0.902 | - | - | - | 0.53 | 3.7 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B0_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B0_infer.tar) | +| PVT_V2_B1 | 0.787 | 0.945 | - | - | - | 2.0 | 14.0 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B1_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B1_infer.tar) | +| PVT_V2_B2 | 0.821 | 0.960 | - | - | - | 3.9 | 25.4 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B2_infer.tar) | +| PVT_V2_B2_Linear | 0.821 | 0.961 | - | - | - | 3.8 | 22.6 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B2_Linear_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B2_Linear_infer.tar) | +| PVT_V2_B3 | 0.831 | 0.965 | - | - |- | 6.7 | 45.2 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B3_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B3_infer.tar) | +| PVT_V2_B4 | 0.836 | 0.967 | - | - | - | 9.8 | 62.6 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B4_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B4_infer.tar) | +| PVT_V2_B5 | 0.837 | 0.966 | - | - | - | 11.4 | 82.0 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/PVT_V2_B5_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PVT_V2_B5_infer.tar) | + + + + +## 25. MobileViT 系列 [[42](#ref42)] + +关于 MobileViT 系列模型的精度、速度指标如下表所示，更多介绍可以参考：[MobileViT 系列模型文档](../models/MobileViT.md)。 + +| 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(M) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | +| ---------- | --------- | --------- | ---------------- | ---------------- | -------- | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | +| MobileViT_XXS | 0.6867 | 0.8878 | - | - | - | 1849.35 | 5.59 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XXS_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileViT_XXS_infer.tar) | +| MobileViT_XS | 0.7454 | 0.9227 | - | - | - | 930.75 | 2.33 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XS_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileViT_XS_infer.tar) | +| MobileViT_S | 0.7814 | 0.9413 | - | - | - | 337.24 | 1.28 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_S_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileViT_S_infer.tar) | + + + +## 26. 其他模型 + +关于 AlexNet [[18](#ref18)]、SqueezeNet 系列 [[19](#ref19)]、VGG 系列 [[20](#ref20)]、DarkNet53 [[21](#ref21)] 等模型的精度、速度指标如下表所示，更多介绍可以参考：[其他模型文档](../models/Others.md)。 | 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | time(ms)
bs=8 | FLOPs(G) | Params(M) | 预训练模型下载地址 | inference模型下载地址 | |------------------------|-----------|-----------|------------------|------------------|----------|-----------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------| @@ -526,3 +562,92 @@ ViT(Vision Transformer) 与 DeiT（Data-efficient Image Transformers）系列模 | VGG16 | 0.720 | 0.907 | 2.48 | 6.79 | 12.33 | 15.470 | 138.35 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG16_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/VGG16_infer.tar) | | VGG19 | 0.726 | 0.909 | 2.93 | 8.28 | 15.21 | 19.63 | 143.66 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG19_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/VGG19_infer.tar) | | DarkNet53 | 0.780 | 0.941 | 2.79 | 6.42 | 10.89 | 9.31 | 41.65 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DarkNet53_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/DarkNet53_infer.tar) | + + + +## 参考文献 + +[1] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778. + +[2] He T, Zhang Z, Zhang H, et al. Bag of tricks for image classification with convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2019: 558-567. + +[3] Howard A, Sandler M, Chu G, et al. Searching for mobilenetv3[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1314-1324. + +[4] Sandler M, Howard A, Zhu M, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 4510-4520. + +[5] Howard A G, Zhu M, Chen B, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[J]. arXiv preprint arXiv:1704.04861, 2017. + +[6] Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 116-131. + +[7] Xie S, Girshick R, Dollár P, et al. Aggregated residual transformations for deep neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1492-1500. + +[8] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141. + +[9] Gao S, Cheng M M, Zhao K, et al. Res2net: A new multi-scale backbone architecture[J]. IEEE transactions on pattern analysis and machine intelligence, 2019. + +[10] Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 1-9. + +[11] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, inception-resnet and the impact of residual connections on learning[C]//Thirty-first AAAI conference on artificial intelligence. 2017. + +[12] Chollet F. Xception: Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1251-1258. + +[13] Wang J, Sun K, Cheng T, et al. Deep high-resolution representation learning for visual recognition[J]. arXiv preprint arXiv:1908.07919, 2019. + +[14] Chen Y, Li J, Xiao H, et al. Dual path networks[C]//Advances in neural information processing systems. 2017: 4467-4475. + +[15] Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 4700-4708. + +[16] Tan M, Le Q V. Efficientnet: Rethinking model scaling for convolutional neural networks[J]. arXiv preprint arXiv:1905.11946, 2019. + +[17] Mahajan D, Girshick R, Ramanathan V, et al. Exploring the limits of weakly supervised pretraining[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 181-196. + +[18] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]//Advances in neural information processing systems. 2012: 1097-1105. + +[19] Iandola F N, Han S, Moskewicz M W, et al. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size[J]. arXiv preprint arXiv:1602.07360, 2016. + +[20] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014. + +[21] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788. + +[22] Ding X, Guo Y, Ding G, et al. Acnet: Strengthening the kernel skeletons for powerful cnn via asymmetric convolution blocks[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1911-1920. + +[23] Han K, Wang Y, Tian Q, et al. GhostNet: More features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 1580-1589. + +[24] Zhang H, Wu C, Zhang Z, et al. Resnest: Split-attention networks[J]. arXiv preprint arXiv:2004.08955, 2020. + +[25] Radosavovic I, Kosaraju R P, Girshick R, et al. Designing network design spaces[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 10428-10436. + +[26] C.Szegedy, V.Vanhoucke, S.Ioffe, J.Shlens, and Z.Wojna. Rethinking the inception architecture for computer vision. arXiv preprint arXiv:1512.00567, 2015. + +[27] Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin and Baining Guo. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. + +[28]Cheng Cui, Tingquan Gao, Shengyu Wei, Yuning Du, Ruoyu Guo, Shuilong Dong, Bin Lu, Ying Zhou, Xueying Lv, Qiwen Liu, Xiaoguang Hu, Dianhai Yu, Yanjun Ma. PP-LCNet: A Lightweight CPU Convolutional Neural Network. + +[29]Mingxing Tan, Quoc V. Le. MixConv: Mixed Depthwise Convolutional Kernels. + +[30]Dongyoon Han, Sangdoo Yun, Byeongho Heo, YoungJoon Yoo. Rethinking Channel Dimensions for Efficient Model Design. + +[31]Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby. AN IMAGE IS WORTH 16X16 WORDS: +TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE. + +[32]Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Herve Jegou. Training data-efficient image transformers & distillation through attention. + +[33]Benjamin Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Herve Jegou, Matthijs Douze. LeViT: a Vision Transformer in ConvNet’s Clothing for Faster Inference. + +[34]Xiangxiang Chu, Zhi Tian, Yuqing Wang, Bo Zhang, Haibing Ren, Xiaolin Wei, Huaxia Xia, Chunhua Shen. Twins: Revisiting the Design of Spatial Attention in Vision Transformers. + +[35]Kai Han, An Xiao, Enhua Wu, Jianyuan Guo, Chunjing Xu, Yunhe Wang. Transformer in Transformer. + +[36]Xiaohan Ding, Xiangyu Zhang, Ningning Ma, Jungong Han, Guiguang Ding, Jian Sun. RepVGG: Making VGG-style ConvNets Great Again. + +[37]Ping Chao, Chao-Yang Kao, Yu-Shan Ruan, Chien-Hsiang Huang, Youn-Long Lin. HarDNet: A Low Memory Traffic Network. + +[38]Fisher Yu, Dequan Wang, Evan Shelhamer, Trevor Darrell. Deep Layer Aggregation. + +[39]Duo Lim Jie Hu, Changhu Wang, Xiangtai Li, Qi She, Lei Zhu, Tong Zhang, Qifeng Chen. Involution: Inverting the Inherence of Convolution for Visual Recognition. + +[40]Xiaoyi Dong, Jianmin Bao, Dongdong Chen, Weiming Zhang, Nenghai Yu, Lu Yuan, Dong Chen, Baining Guo. CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows. + +[41]Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, Ling Shao. PVTv2: Improved Baselines with Pyramid Vision Transformer. + +[42]Sachin Mehta, Mohammad Rastegari. MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer. diff --git a/docs/zh_CN/faq_series/faq_2020_s1.md b/docs/zh_CN/faq_series/faq_2020_s1.md index e780a54a993878878b2377696d5478ec94f6ae7d..09a1cda7e48a3e5c30d32d6f2486b9bd208b73a5 100644 --- a/docs/zh_CN/faq_series/faq_2020_s1.md +++ b/docs/zh_CN/faq_series/faq_2020_s1.md @@ -122,14 +122,15 @@ ResNet 系列模型中，相比于其他模型，ResNet_vd 模型在预测速度 **A**： -* 可以使用自动混合精度进行训练，这在精度几乎无损的情况下，可以有比较明显的速度收益，以 ResNet50 为例，PaddleClas 中使用自动混合精度训练的配置文件可以参考：[ResNet50_fp16.yml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_fp16.yaml)，主要就是需要在标准的配置文件中添加以下几行 +* 可以使用自动混合精度进行训练，这在精度几乎无损的情况下，可以有比较明显的速度收益，以 ResNet50 为例，PaddleClas 中使用自动混合精度训练的配置文件可以参考：[ResNet50_amp_O1.yml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml)，主要就是需要在标准的配置文件中添加以下几行 -``` +```yaml # mixed precision training AMP: scale_loss: 128.0 use_dynamic_loss_scaling: True - use_pure_fp16: &use_pure_fp16 True + # O1: mixed fp16 + level: O1 ``` * 可以开启 dali，将数据预处理方法放在 GPU 上运行，在模型比较小时（reader 耗时占比更高一些），开启 dali 会带来比较明显的训练速度收益，在训练的时候，添加 `-o Global.use_dali=True` 即可使用 dali 进行训练，更多关于 dali 安装与介绍可以参考：[dali 安装教程](https://docs.nvidia.com/deeplearning/dali/user-guide/docs/installation.html#nightly-builds)。 diff --git a/docs/zh_CN/faq_series/faq_selected_30.md b/docs/zh_CN/faq_series/faq_selected_30.md index 33826738097c20a154b15c57f66ef6c1cecc61b0..10d2d89d912805b472321a9263bb7b11ebe16ace 100644 --- a/docs/zh_CN/faq_series/faq_selected_30.md +++ b/docs/zh_CN/faq_series/faq_selected_30.md @@ -31,7 +31,8 @@ >> * Q: 怎样根据自己的任务选择合适的模型进行训练？ -* A: 如果希望在服务器部署，或者希望精度尽可能地高，对模型存储大小或者预测速度的要求不是很高，那么推荐使用 ResNet_vd、Res2Net_vd、DenseNet、Xception 等适合于服务器端的系列模型；如果希望在移动端侧部署，则推荐使用 MobileNetV3、GhostNet 等适合于移动端的系列模型。同时，我们推荐在选择模型的时候可以参考[模型库](../models/models_intro.md)中的速度-精度指标图。 +* A: 如果希望在服务器部署，或者希望精度尽可能地高，对模型存储大小或者预测速度的要求不是很高，那么推荐使用 ResNet_vd、Res2Net_vd、DenseNet、Xception 等适合于服务器端的系列模型；如果希望在移动端侧部署，则推荐使用 MobileNetV3、GhostNet + 等适合于移动端的系列模型。同时，我们推荐在选择模型的时候可以参考[模型库](../algorithm_introduction/ImageNet_models.md)中的速度-精度指标图。 >> * Q: 如何进行参数初始化，什么样的初始化可以加快模型收敛？ @@ -232,11 +233,13 @@ Loss: * A: 如果希望使用 TensorRT 进行模型预测推理的话，需要安装或是自己编译带 TensorRT 的 PaddlePaddle，Linux、Windows、macOS 系统的用户下载安装可以参考参考[下载预测库](https://paddleinference.paddlepaddle.org.cn/user_guides/download_lib.html)，如果没有符合您所需要的版本，则需要本地编译安装，编译方法可以参考[源码编译](https://paddleinference.paddlepaddle.org.cn/user_guides/source_compile.html)。 >> * Q: 怎样在训练的时候使用自动混合精度(Automatic Mixed Precision, AMP)训练呢？ -* A: 可以参考 [ResNet50_fp16.yaml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_fp16.yaml) 这个配置文件；具体地，如果希望自己的配置文件在模型训练的时候也支持自动混合精度，可以在配置文件中添加下面的配置信息。 -``` +* A: 可以参考 [ResNet50_amp_O1.yaml](../../../ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml) 这个配置文件；具体地，如果希望自己的配置文件在模型训练的时候也支持自动混合精度，可以在配置文件中添加下面的配置信息。 + +```yaml # mixed precision training AMP: scale_loss: 128.0 use_dynamic_loss_scaling: True - use_pure_fp16: &use_pure_fp16 True + # O1: mixed fp16 + level: O1 ``` diff --git a/docs/zh_CN/inference_deployment/paddle_hub_serving_deploy.md b/docs/zh_CN/inference_deployment/paddle_hub_serving_deploy.md index 1589bc6a5a9b7cae1818e7eeb3edd9f4d98719bb..e3892e9a96810c418ec508a555a9d276b3ba73ae 100644 --- a/docs/zh_CN/inference_deployment/paddle_hub_serving_deploy.md +++ b/docs/zh_CN/inference_deployment/paddle_hub_serving_deploy.md @@ -15,8 +15,8 @@ PaddleClas 支持通过 PaddleHub 快速进行服务化部署。目前支持图 - [5.2 配置文件启动](#5.2) - [6. 发送预测请求](#6) - [7. 自定义修改服务模块](#7) - - + + ## 1. 简介 @@ -55,7 +55,7 @@ pip3 install paddlehub==2.1.0 --upgrade -i https://pypi.tuna.tsinghua.edu.cn/sim ``` 需要注意， * 模型文件（包括 `.pdmodel` 与 `.pdiparams`）名称必须为 `inference`。 - * 我们也提供了大量基于 ImageNet-1k 数据集的预训练模型，模型列表及下载地址详见[模型库概览](../models/models_intro.md)，也可以使用自己训练转换好的模型。 + * 我们也提供了大量基于 ImageNet-1k 数据集的预训练模型，模型列表及下载地址详见[模型库概览](../algorithm_introduction/ImageNet_models.md)，也可以使用自己训练转换好的模型。 diff --git a/docs/zh_CN/inference_deployment/paddle_lite_deploy.md b/docs/zh_CN/inference_deployment/paddle_lite_deploy.md index 25cf5a2c6f602d8da0a160b0c650fd32ccc9f754..68480f769a67aae33ca614b0eede2581fcf57392 100644 --- a/docs/zh_CN/inference_deployment/paddle_lite_deploy.md +++ b/docs/zh_CN/inference_deployment/paddle_lite_deploy.md @@ -1,11 +1,8 @@ -# PaddleLite 推理部署 ---- - -本教程将介绍基于[Paddle Lite](https://github.com/PaddlePaddle/Paddle-Lite)在移动端部署 PaddleClas 分类模型的详细步骤。识别模型的部署将在近期支持，敬请期待。 +# 端侧部署 -Paddle Lite 是飞桨轻量化推理引擎，为手机、IOT 端提供高效推理能力，并广泛整合跨平台硬件，为端侧部署及应用落地问题提供轻量化的部署方案。 +本教程将介绍基于[Paddle Lite](https://github.com/PaddlePaddle/Paddle-Lite) 在移动端部署PaddleClas分类模型的详细步骤。 -如果希望直接测试速度，可以参考[Paddle-Lite 移动端 benchmark 测试教程](../others/paddle_mobile_inference.md)。 +Paddle Lite是飞桨轻量化推理引擎，为手机、IOT端提供高效推理能力，并广泛整合跨平台硬件，为端侧部署及应用落地问题提供轻量化的部署方案。如果希望直接测试速度，可以参考[Paddle-Lite移动端benchmark测试教程](../../docs/zh_CN/extension/paddle_mobile_inference.md)。 --- @@ -18,53 +15,54 @@ Paddle Lite 是飞桨轻量化推理引擎，为手机、IOT 端提供高效推 - [2.1.1 pip 安装 paddlelite 并进行转换](#2.1.1) - [2.1.2 源码编译 Paddle-Lite 生成 opt 工具](#2.1.2) - [2.1.3 转换示例](#2.1.3) + - [2.1.4 执行编译，得到可执行文件clas_system](#2.1.4) - [2.2 与手机联调](#2.2) - [3. FAQ](#3) - ## 1. 准备环境 -Paddle Lite 目前支持以下平台部署： -* 电脑（编译 Paddle Lite） -* 安卓手机（armv7 或 armv8） +### 运行准备 +- 电脑（编译Paddle Lite） +- 安卓手机（armv7或armv8） - ### 1.1 准备交叉编译环境 +交叉编译环境用于编译 Paddle Lite 和 PaddleClas 的C++ demo。 +支持多种开发环境，不同开发环境的编译流程请参考对应文档。 -交叉编译环境用于编译 Paddle Lite 和 PaddleClas 的 C++ demo。 -支持多种开发环境，关于 Docker、Linux、macOS、Windows 等不同开发环境的编译流程请参考[文档](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html)。 +1. [Docker](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#docker) +2. [Linux](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#linux) +3. [MAC OS](https://paddle-lite.readthedocs.io/zh/latest/source_compile/compile_env.html#mac-os) - ### 1.2 准备预测库 预测库有两种获取方式： 1. [建议]直接下载，预测库下载链接如下： |平台|预测库下载链接| |-|-| - |Android|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv7.gcc.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/Android/gcc/inference_lite_lib.android.armv8.gcc.c++_static.with_extra.with_cv.tar.gz)| - |iOS|[arm7](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://paddlelite-data.bj.bcebos.com/Release/2.8-rc/iOS/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| + |Android|[arm7](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.android.armv7.clang.c++_static.with_extra.with_cv.tar.gz) / [arm8](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv.tar.gz)| + |iOS|[arm7](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.ios.armv7.with_cv.with_extra.tiny_publish.tar.gz) / [arm8](https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.ios.armv8.with_cv.with_extra.tiny_publish.tar.gz)| **注**： 1. 如果是从 Paddle-Lite [官方文档](https://paddle-lite.readthedocs.io/zh/latest/quick_start/release_lib.html#android-toolchain-gcc)下载的预测库， - 注意选择 `with_extra=ON，with_cv=ON` 的下载链接。 - 2. 如果使用量化的模型部署在端侧，建议使用 Paddle-Lite develop 分支编译预测库。 - -2. 编译 Paddle-Lite 得到预测库，Paddle-Lite 的编译方式如下： + 注意选择`with_extra=ON，with_cv=ON`的下载链接。 + 2. 如果使用量化的模型部署在端侧，建议使用Paddle-Lite develop分支编译预测库。 +2. 编译Paddle-Lite得到预测库，Paddle-Lite的编译方式如下： ```shell git clone https://github.com/PaddlePaddle/Paddle-Lite.git cd Paddle-Lite -# 如果使用编译方式，建议使用 develop 分支编译预测库 +# 如果使用编译方式，建议使用develop分支编译预测库 git checkout develop ./lite/tools/build_android.sh --arch=armv8 --with_cv=ON --with_extra=ON ``` -**注意**：编译 Paddle-Lite 获得预测库时，需要打开`--with_cv=ON --with_extra=ON` 两个选项，`--arch` 表示 `arm` 版本，这里指定为 armv8，更多编译命令介绍请参考[Linux x86 环境下编译适用于 Android 的库](https://paddle-lite.readthedocs.io/zh/latest/source_compile/linux_x86_compile_android.html)，关于其他平台的编译操作，具体请参考[PaddleLite](https://paddle-lite.readthedocs.io/zh/latest/)中`源码编译`部分。 -直接下载预测库并解压后，可以得到 `inference_lite_lib.android.armv8/`文件夹，通过编译 Paddle-Lite 得到的预测库位于 `Paddle-Lite/build.lite.android.armv8.gcc/inference_lite_lib.android.armv8/`文件夹下。 +**注意**：编译Paddle-Lite获得预测库时，需要打开`--with_cv=ON --with_extra=ON`两个选项，`--arch`表示`arm`版本，这里指定为armv8，更多编译命令介绍请参考[链接](https://paddle-lite.readthedocs.io/zh/latest/user_guides/Compile/Android.html#id2)。 + +直接下载预测库并解压后，可以得到`inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/`文件夹，通过编译Paddle-Lite得到的预测库位于`Paddle-Lite/build.lite.android.armv8.gcc/inference_lite_lib.android.armv8/`文件夹下。 预测库的文件目录如下： ``` -inference_lite_lib.android.armv8/ +inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/ |-- cxx C++ 预测库和头文件 | |-- include C++ 头文件 | | |-- paddle_api.h @@ -77,7 +75,7 @@ inference_lite_lib.android.armv8/ | `-- lib C++预测库 | |-- libpaddle_api_light_bundled.a C++静态库 | `-- libpaddle_light_api_shared.so C++动态库 -|-- java Java 预测库 +|-- java Java预测库 | |-- jar | | `-- PaddlePredictor.jar | |-- so @@ -88,47 +86,43 @@ inference_lite_lib.android.armv8/ | `-- java Java 预测库demo ``` - -## 2. 开始运行 +## 2 开始运行 - ### 2.1 模型优化 -Paddle-Lite 提供了多种策略来自动优化原始的模型，其中包括量化、子图融合、混合精度、Kernel 优选等方法，使用 Paddle-Lite 的 `opt` 工具可以自动对 inference 模型进行优化，目前支持两种优化方式，优化后的模型更轻量，模型运行速度更快。在进行模型优化前，需要先准备 `opt` 优化工具，有以下两种方式。 +Paddle-Lite 提供了多种策略来自动优化原始的模型，其中包括量化、子图融合、混合调度、Kernel优选等方法，使用Paddle-Lite的`opt`工具可以自动对inference模型进行优化，目前支持两种优化方式，优化后的模型更轻量，模型运行速度更快。 **注意**：如果已经准备好了 `.nb` 结尾的模型文件，可以跳过此步骤。 - -#### 2.1.1 [建议]pip 安装 paddlelite 并进行转换 +#### 2.1.1 [建议]pip安装paddlelite并进行转换 -Python 下安装 `paddlelite`，目前最高支持 `Python3.7`。 -**注意**：`paddlelite` whl 包版本必须和预测库版本对应。 +Python下安装 `paddlelite`，目前最高支持`Python3.7`。 +**注意**：`paddlelite`whl包版本必须和预测库版本对应。 ```shell -pip install paddlelite==2.8 +pip install paddlelite==2.10 ``` -之后使用 `paddle_lite_opt` 工具可以进行 inference 模型的转换。`paddle_lite_opt` 的部分参数如下 +之后使用`paddle_lite_opt`工具可以进行inference模型的转换。`paddle_lite_opt`的部分参数如下 |选项|说明| |-|-| -|--model_dir|待优化的 PaddlePaddle 模型（非 combined 形式）的路径| -|--model_file|待优化的 PaddlePaddle 模型（combined 形式）的网络结构文件路径| -|--param_file|待优化的 PaddlePaddle 模型（combined 形式）的权重文件路径| -|--optimize_out_type|输出模型类型，目前支持两种类型：protobuf 和 naive_buffer，其中 naive_buffer 是一种更轻量级的序列化/反序列化实现。若您需要在 mobile 端执行模型预测，请将此选项设置为 naive_buffer。默认为 protobuf| +|--model_dir|待优化的PaddlePaddle模型（非combined形式）的路径| +|--model_file|待优化的PaddlePaddle模型（combined形式）的网络结构文件路径| +|--param_file|待优化的PaddlePaddle模型（combined形式）的权重文件路径| +|--optimize_out_type|输出模型类型，目前支持两种类型：protobuf和naive_buffer，其中naive_buffer是一种更轻量级的序列化/反序列化实现。若您需要在mobile端执行模型预测，请将此选项设置为naive_buffer。默认为protobuf| |--optimize_out|优化模型的输出路径| -|--valid_targets|指定模型可执行的 backend，默认为 arm。目前可支持 x86、arm、opencl、npu、xpu，可以同时指定多个 backend（以空格分隔），Model Optimize Tool 将会自动选择最佳方式。如果需要支持华为 NPU（Kirin 810/990 Soc 搭载的达芬奇架构 NPU），应当设置为 npu, arm| -|--record_tailoring_info|当使用 根据模型裁剪库文件 功能时，则设置该选项为 true，以记录优化后模型含有的 kernel 和 OP 信息，默认为 false| +|--valid_targets|指定模型可执行的backend，默认为arm。目前可支持x86、arm、opencl、npu、xpu，可以同时指定多个backend(以空格分隔)，Model Optimize Tool将会自动选择最佳方式。如果需要支持华为NPU（Kirin 810/990 Soc搭载的达芬奇架构NPU），应当设置为npu, arm| +|--record_tailoring_info|当使用 根据模型裁剪库文件 功能时，则设置该选项为true，以记录优化后模型含有的kernel和OP信息，默认为false| -`--model_file` 表示 inference 模型的 model 文件地址，`--param_file` 表示 inference 模型的 param 文件地址；`optimize_out` 用于指定输出文件的名称（不需要添加 `.nb` 的后缀）。直接在命令行中运行 `paddle_lite_opt`，也可以查看所有参数及其说明。 +`--model_file`表示inference模型的model文件地址，`--param_file`表示inference模型的param文件地址；`optimize_out`用于指定输出文件的名称（不需要添加`.nb`的后缀）。直接在命令行中运行`paddle_lite_opt`，也可以查看所有参数及其说明。 - -#### 2.1.2 源码编译 Paddle-Lite 生成 opt 工具 -模型优化需要 Paddle-Lite 的 `opt` 可执行文件，可以通过编译 Paddle-Lite 源码获得，编译步骤如下： +#### 2.1.2 源码编译Paddle-Lite生成opt工具 +模型优化需要Paddle-Lite的`opt`可执行文件，可以通过编译Paddle-Lite源码获得，编译步骤如下： ```shell -# 如果准备环境时已经 clone 了 Paddle-Lite，则不用重新 clone Paddle-Lite +# 如果准备环境时已经clone了Paddle-Lite，则不用重新clone Paddle-Lite git clone https://github.com/PaddlePaddle/Paddle-Lite.git cd Paddle-Lite git checkout develop @@ -136,146 +130,137 @@ git checkout develop ./lite/tools/build.sh build_optimize_tool ``` -编译完成后，`opt` 文件位于 `build.opt/lite/api/` 下，可通过如下方式查看 `opt` 的运行选项和使用方式： - +编译完成后，`opt`文件位于`build.opt/lite/api/`下，可通过如下方式查看`opt`的运行选项和使用方式； ```shell cd build.opt/lite/api/ ./opt ``` -`opt` 的使用方式与参数与上面的 `paddle_lite_opt` 完全一致。 +`opt`的使用方式与参数与上面的`paddle_lite_opt`完全一致。 + #### 2.1.3 转换示例 -下面以 PaddleClas 的 `MobileNetV3_large_x1_0` 模型为例，介绍使用 `paddle_lite_opt` 完成预训练模型到 inference 模型，再到 Paddle-Lite 优化模型的转换。 +下面以PaddleClas的 `MobileNetV3_large_x1_0` 模型为例，介绍使用`paddle_lite_opt`完成预训练模型到inference模型，再到Paddle-Lite优化模型的转换。 ```shell -# 进入 PaddleClas 根目录 +# 进入PaddleClas根目录 cd PaddleClas_root_path -# 下载并解压 inference 模型 +# 下载并解压inference模型 wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/MobileNetV3_large_x1_0_infer.tar tar -xf MobileNetV3_large_x1_0_infer.tar -# 将 inference 模型转化为 Paddle-Lite 优化模型 +# 将inference模型转化为Paddle-Lite优化模型 paddle_lite_opt --model_file=./MobileNetV3_large_x1_0_infer/inference.pdmodel --param_file=./MobileNetV3_large_x1_0_infer/inference.pdiparams --optimize_out=./MobileNetV3_large_x1_0 ``` -最终在当前文件夹下生成 `MobileNetV3_large_x1_0.nb` 的文件。 +最终在当前文件夹下生成`MobileNetV3_large_x1_0.nb`的文件。 + +**注意**：`--optimize_out` 参数为优化后模型的保存路径，无需加后缀`.nb`；`--model_file` 参数为模型结构信息文件的路径，`--param_file` 参数为模型权重信息文件的路径，请注意文件名。 + + + +#### 2.1.4 执行编译，得到可执行文件clas_system + +```shell +# 克隆 Autolog 代码库，以便获取自动化日志 +cd PaddleClas_root_path +cd deploy/lite/ +git clone https://github.com/LDOUBLEV/AutoLog.git +``` + +```shell +# 编译 +make -j +``` -**注意**：`--optimize_out` 参数为优化后模型的保存路径，无需加后缀 `.nb`；`--model_file` 参数为模型结构信息文件的路径，`--param_file` 参数为模型权重信息文件的路径，请注意文件名。 +执行 `make` 命令后，会在当前目录生成 `clas_system` 可执行文件，该文件用于 Lite 预测。 - + ### 2.2 与手机联调 首先需要进行一些准备工作。 -1. 准备一台 arm8 的安卓手机，如果编译的预测库和 opt 文件是 armv7，则需要 arm7 的手机，并修改 Makefile 中 `ARM_ABI = arm7`。 -2. 电脑上安装 ADB 工具，用于调试。 ADB 安装方式如下： +1. 准备一台arm8的安卓手机，如果编译的预测库和opt文件是armv7，则需要arm7的手机，并修改Makefile中`ARM_ABI = arm7`。 +2. 电脑上安装ADB工具，用于调试。 ADB安装方式如下： + + 3.1. MAC电脑安装ADB: - * MAC 电脑安装 ADB: ```shell brew cask install android-platform-tools ``` - * Linux 安装 ADB + 3.2. Linux安装ADB ```shell sudo apt update sudo apt install -y wget adb ``` - * Window 安装 ADB - win 上安装需要去谷歌的安卓平台下载 ADB 软件包进行安装：[链接](https://developer.android.com/studio) + 3.3. Window安装ADB + + win上安装需要去谷歌的安卓平台下载ADB软件包进行安装：[链接](https://developer.android.com/studio) -3. 手机连接电脑后，开启手机 `USB 调试` 选项，选择 `文件传输` 模式，在电脑终端中输入： +3. 手机连接电脑后，开启手机`USB调试`选项，选择`文件传输`模式，在电脑终端中输入： ```shell adb devices ``` -如果有 device 输出，则表示安装成功，如下所示： +如果有device输出，则表示安装成功，如下所示： ``` List of devices attached 744be294 device ``` -4. 准备优化后的模型、预测库文件、测试图像和类别映射文件。 -```shell -cd PaddleClas_root_path -cd deploy/lite/ - -# 运行 prepare.sh -# prepare.sh 会将预测库文件、测试图像和使用的字典文件放置在预测库中的 demo/cxx/clas 文件夹下 -sh prepare.sh /{lite prediction library path}/inference_lite_lib.android.armv8 - -# 进入 lite demo 的工作目录 -cd /{lite prediction library path}/inference_lite_lib.android.armv8/ -cd demo/cxx/clas/ +4. 将优化后的模型、预测库文件、测试图像和类别映射文件push到手机上。 -# 将 C++ 预测动态库 so 文件复制到 debug 文件夹中 -cp ../../../cxx/lib/libpaddle_light_api_shared.so ./debug/ -``` - -`prepare.sh` 以 `PaddleClas/deploy/lite/imgs/tabby_cat.jpg` 作为测试图像，将测试图像复制到 `demo/cxx/clas/debug/` 文件夹下。 -将 `paddle_lite_opt` 工具优化后的模型文件放置到 `/{lite prediction library path}/inference_lite_lib.android.armv8/demo/cxx/clas/debug/` 文件夹下。本例中，使用 [2.1.3 转换示例](#2.1.3) 生成的 `MobileNetV3_large_x1_0.nb` 模型文件。 - -执行完成后，clas 文件夹下将有如下文件格式： - -``` -demo/cxx/clas/ -|-- debug/ -| |--MobileNetV3_large_x1_0.nb 优化后的分类器模型文件 -| |--tabby_cat.jpg 待测试图像 -| |--imagenet1k_label_list.txt 类别映射文件 -| |--libpaddle_light_api_shared.so C++预测库文件 -| |--config.txt 分类预测超参数配置 -|-- config.txt 分类预测超参数配置 -|-- image_classfication.cpp 图像分类代码文件 -|-- Makefile 编译文件 +```shell +adb shell mkdir -p /data/local/tmp/arm_cpu/ +adb push clas_system /data/local/tmp/arm_cpu/ +adb shell chmod +x /data/local/tmp/arm_cpu//clas_system +adb push inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv/cxx/lib/libpaddle_light_api_shared.so /data/local/tmp/arm_cpu/ +adb push MobileNetV3_large_x1_0.nb /data/local/tmp/arm_cpu/ +adb push config.txt /data/local/tmp/arm_cpu/ +adb push ../../ppcls/utils/imagenet1k_label_list.txt /data/local/tmp/arm_cpu/ +adb push imgs/tabby_cat.jpg /data/local/tmp/arm_cpu/ ``` #### 注意： -* 上述文件中，`imagenet1k_label_list.txt` 是 ImageNet1k 数据集的类别映射文件，如果使用自定义的类别，需要更换该类别映射文件。 +* 上述文件中，`imagenet1k_label_list.txt` 是ImageNet1k数据集的类别映射文件，如果使用自定义的类别，需要更换该类别映射文件。 -* `config.txt` 包含了分类器的超参数，如下： +* `config.txt` 包含了分类器的超参数，如下： ```shell clas_model_file ./MobileNetV3_large_x1_0.nb # 模型文件地址 -label_path ./imagenet1k_label_list.txt # 类别映射文本文件 -resize_short_size 256 # resize 之后的短边边长 -crop_size 224 # 裁剪后用于预测的边长 -visualize 0 # 是否进行可视化，如果选择的话，会在当前文件夹下生成名为 clas_result.png 的图像文件。 +label_path ./imagenet1k_label_list.txt # 类别映射文本文件 +resize_short_size 256 # resize之后的短边边长 +crop_size 224 # 裁剪后用于预测的边长 +visualize 0 # 是否进行可视化，如果选择的话，会在当前文件夹下生成名为clas_result.png的图像文件 +num_threads 1 # 线程数，默认是1。 +precision FP32 # 精度类型，可以选择 FP32 或者 INT8，默认是 FP32。 +runtime_device arm_cpu # 设备类型，默认是 arm_cpu +enable_benchmark 0 # 是否开启benchmark， 默认是 0 +tipc_benchmark 0 # 是否开启tipc_benchmark，默认是 0 ``` -5. 启动调试，上述步骤完成后就可以使用 ADB 将文件夹 `debug/` push 到手机上运行，步骤如下： - -```shell -# 执行编译，得到可执行文件 clas_system -make -j - -# 将编译得到的可执行文件移动到 debug 文件夹中 -mv clas_system ./debug/ +5. 执行预测命令 -# 将上述 debug 文件夹 push 到手机上 -adb push debug /data/local/tmp/ +执行以下命令，可完成在手机上的预测。 -adb shell -cd /data/local/tmp/debug -export LD_LIBRARY_PATH=/data/local/tmp/debug:$LD_LIBRARY_PATH - -# clas_system 可执行文件的使用方式为: -# ./clas_system 配置文件路径 测试图像路径 -./clas_system ./config.txt ./tabby_cat.jpg +```shell +adb shell 'export LD_LIBRARY_PATH=/data/local/tmp/arm_cpu/; /data/local/tmp/arm_cpu/clas_system /data/local/tmp/arm_cpu/config.txt /data/local/tmp/arm_cpu/tabby_cat.jpg' ``` -如果对代码做了修改，则需要重新编译并 push 到手机上。 - 运行效果如下： -![](../../images/inference_deployment/lite_demo_result.png) +
+ +
- -## 3. FAQ + +## FAQ Q1：如果想更换模型怎么办，需要重新按照流程走一遍吗？ A1：如果已经走通了上述步骤，更换模型只需要替换 `.nb` 模型文件即可，同时要注意修改下配置文件中的 `.nb` 文件路径以及类别映射文件（如有必要）。 Q2：换一个图测试怎么做？ A2：替换 debug 下的测试图像为你想要测试的图像，使用 ADB 再次 push 到手机上即可。 + diff --git a/docs/zh_CN/models/CSWinTransformer.md b/docs/zh_CN/models/CSWinTransformer.md new file mode 100644 index 0000000000000000000000000000000000000000..21631392092a9b13ecbcc06b6bbbccf34ad10439 --- /dev/null +++ b/docs/zh_CN/models/CSWinTransformer.md @@ -0,0 +1,24 @@ +# CSWinTransformer +--- +## 目录 + +* [1. 概述](#1) +* [2. 精度、FLOPs 和参数量](#2) + + + +## 1. 概述 +CSWinTransformer 是一种新的视觉 Transformer 网络，可以用作计算机视觉领域的通用骨干网路。 CSWinTransformer 提出了通过十字形的窗口来做 self-attention，它不仅计算效率非常高，而且能够通过两层计算就获得全局的感受野。CSWinTransformer 还提出了新的编码方式：LePE，进一步提高了模型的准确率。[论文地址](https://arxiv.org/abs/2107.00652)。 + + + +## 2. 精度、FLOPs 和参数量 + +| Models | Top1 | Top5 | Reference
top1 | Reference
top5 | FLOPs
(G) | Params
(M) | +|:--:|:--:|:--:|:--:|:--:|:--:|:--:| +| CSWinTransformer_tiny_224 | 0.8281 | 0.9628 | 0.828 | - | 4.1 | 22 | +| CSWinTransformer_small_224 | 0.8358 | 0.9658 | 0.836 | - | 6.4 | 35 | +| CSWinTransformer_base_224 | 0.8420 | 0.9692 | 0.842 | - | 14.3 | 77 | +| CSWinTransformer_large_224 | 0.8643 | 0.9799 | 0.865 | - | 32.2 | 173.3 | +| CSWinTransformer_base_384 | 0.8550 | 0.9749 | 0.855 | - | 42.2 | 77 | +| CSWinTransformer_large_384 | 0.8748 | 0.9833 | 0.875 | - | 94.7 | 173.3 | diff --git a/docs/zh_CN/models/MobileViT.md b/docs/zh_CN/models/MobileViT.md new file mode 100644 index 0000000000000000000000000000000000000000..2980fb38f80c412a18b73674eac7cd3fd7793ce5 --- /dev/null +++ b/docs/zh_CN/models/MobileViT.md @@ -0,0 +1,22 @@ +# MobileviT +--- +## 目录 + +* [1. 概述](#1) +* [2. 精度、FLOPs 和参数量](#2) + + + +## 1. 概述 + +MobileViT 是一个轻量级的视觉 Transformer 网络，可以用作计算机视觉领域的通用骨干网路。 MobileViT 结合了 CNN 和 Transformer 的优势，可以更好的处理全局特征和局部特征，更好地解决 Transformer 模型缺乏归纳偏置的问题，最终，在同样参数量下，与其他 SOTA 模型相比，在图像分类、目标检测、语义分割任务上都有大幅提升。[论文地址](https://arxiv.org/pdf/2110.02178.pdf)。 + + + +## 2. 精度、FLOPs 和参数量 + +| Models | Top1 | Top5 | Reference
top1 | Reference
top5 | FLOPs
(M) | Params
(M) | +|:--:|:--:|:--:|:--:|:--:|:--:|:--:| +| MobileViT_XXS | 0.6867 | 0.8878 | 0.690 | - | 1849.35 | 5.59 | +| MobileViT_XS | 0.7454 | 0.9227 | 0.747 | - | 930.75 | 2.33 | +| MobileViT_S | 0.7814 | 0.9413 | 0.783 | - | 337.24 | 1.28 | diff --git a/docs/zh_CN/models/models_intro.md b/docs/zh_CN/models/models_intro.md deleted file mode 100644 index f529307f7af85f049f128cce09fe484ae21174e8..0000000000000000000000000000000000000000 --- a/docs/zh_CN/models/models_intro.md +++ /dev/null @@ -1,406 +0,0 @@ -# 模型库概览 ---- -## 目录 - -* [1. 概述](#1) -* [2. 评估环境](#2) -* [3. 预训练模型列表及下载地址](#3) -* [4. 参考文献](#4) - - - -## 1. 概述 - -基于 ImageNet1k 分类数据集，PaddleClas 支持的 36 种系列分类网络结构以及对应的 175 个图像分类预训练模型如下所示，训练技巧、每个系列网络结构的简单介绍和性能评估将在相应章节展现。 - - - -## 2. 评估环境 -* Arm 的评估环境基于骁龙 855（SD855）。 -* Intel CPU 的评估环境基于 Intel(R) Xeon(R) Gold 6148。 -* GPU 评估环境基于 V100 和 TensorRT。 - -![](../../images/models/V100_benchmark/v100.fp32.bs1.main_fps_top1_s.png) - -![](../../images/models/V100_benchmark/v100.fp32.bs1.visiontransformer.png) - -> 如果您觉得此文档对您有帮助，欢迎 star 我们的项目：[https://github.com/PaddlePaddle/PaddleClas](https://github.com/PaddlePaddle/PaddleClas) - - - -## 3. 预训练模型列表及下载地址 - -- ResNet 及其 Vd 系列 - - ResNet 系列[[1](#ref1)]([论文地址](http://openaccess.thecvf.com/content_cvpr_2016/html/He_Deep_Residual_Learning_CVPR_2016_paper.html)) - - [ResNet18](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet18_pretrained.pdparams) - - [ResNet34](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet34_pretrained.pdparams) - - [ResNet50](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_pretrained.pdparams) - - [ResNet101](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet101_pretrained.pdparams) - - [ResNet152](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet152_pretrained.pdparams) - - ResNet_vc、ResNet_vd 系列[[2](#ref2)]([论文地址](https://arxiv.org/abs/1812.01187)) - - [ResNet50_vc](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vc_pretrained.pdparams) - - [ResNet18_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet18_vd_pretrained.pdparams) - - [ResNet34_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet34_vd_pretrained.pdparams) - - [ResNet34_vd_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet34_vd_ssld_pretrained.pdparams) - - [ResNet50_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_pretrained.pdparams) - - [ResNet50_vd_v2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_v2_pretrained.pdparams) - - [ResNet101_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet101_vd_pretrained.pdparams) - - [ResNet152_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet152_vd_pretrained.pdparams) - - [ResNet200_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet200_vd_pretrained.pdparams) - - [ResNet50_vd_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_ssld_pretrained.pdparams) - - [Fix_ResNet50_vd_ssld_v2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Fix_ResNet50_vd_ssld_v2_pretrained.pdparams) - - [ResNet101_vd_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet101_vd_ssld_pretrained.pdparams) - - -- 轻量级模型系列 - - PP-LCNet 系列[[28](#ref28)]([论文地址](https://arxiv.org/pdf/2109.15099.pdf)) - - [PPLCNet_x0_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_25_pretrained.pdparams) - - [PPLCNet_x0_35](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_35_pretrained.pdparams) - - [PPLCNet_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_pretrained.pdparams) - - [PPLCNet_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_75_pretrained.pdparams) - - [PPLCNet_x1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_pretrained.pdparams) - - [PPLCNet_x1_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_5_pretrained.pdparams) - - [PPLCNet_x2_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_0_pretrained.pdparams) - - [PPLCNet_x2_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_pretrained.pdparams) - - [PPLCNet_x0_5_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_ssld_pretrained.pdparams) - - [PPLCNet_x1_0_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_ssld_pretrained.pdparams) - - [PPLCNet_x2_5_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_ssld_pretrained.pdparams) - - MobileNetV3 系列[[3](#ref3)]([论文地址](https://arxiv.org/abs/1905.02244)) - - [MobileNetV3_large_x0_35](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_35_pretrained.pdparams) - - [MobileNetV3_large_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams) - - [MobileNetV3_large_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_75_pretrained.pdparams) - - [MobileNetV3_large_x1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x1_0_pretrained.pdparams) - - [MobileNetV3_large_x1_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x1_25_pretrained.pdparams) - - [MobileNetV3_small_x0_35](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x0_35_pretrained.pdparams) - - [MobileNetV3_small_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x0_5_pretrained.pdparams) - - [MobileNetV3_small_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x0_75_pretrained.pdparams) - - [MobileNetV3_small_x1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x1_0_pretrained.pdparams) - - [MobileNetV3_small_x1_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x1_25_pretrained.pdparams) - - [MobileNetV3_large_x1_0_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x1_0_ssld_pretrained.pdparams) - - [MobileNetV3_large_x1_0_ssld_int8]()(coming soon) - - [MobileNetV3_small_x1_0_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_small_x1_0_ssld_pretrained.pdparams) - - MobileNetV2 系列[[4](#ref4)]([论文地址](https://arxiv.org/abs/1801.04381)) - - [MobileNetV2_x0_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_25_pretrained.pdparams) - - [MobileNetV2_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_5_pretrained.pdparams) - - [MobileNetV2_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_75_pretrained.pdparams) - - [MobileNetV2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_pretrained.pdparams) - - [MobileNetV2_x1_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x1_5_pretrained.pdparams) - - [MobileNetV2_x2_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x2_0_pretrained.pdparams) - - [MobileNetV2_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_ssld_pretrained.pdparams) - - MobileNetV1 系列[[5](#ref5)]([论文地址](https://arxiv.org/abs/1704.04861)) - - [MobileNetV1_x0_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_x0_25_pretrained.pdparams) - - [MobileNetV1_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_x0_5_pretrained.pdparams) - - [MobileNetV1_x0_75](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_x0_75_pretrained.pdparams) - - [MobileNetV1](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_pretrained.pdparams) - - [MobileNetV1_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV1_ssld_pretrained.pdparams) - - ShuffleNetV2 系列[[6](#ref6)]([论文地址](https://arxiv.org/abs/1807.11164)) - - [ShuffleNetV2_x0_25](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams) - - [ShuffleNetV2_x0_33](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams) - - [ShuffleNetV2_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams) - - [ShuffleNetV2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams) - - [ShuffleNetV2_x1_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams) - - [ShuffleNetV2_x2_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams) - - [ShuffleNetV2_swish](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams) - - GhostNet 系列[[23](#ref23)]([论文地址](https://arxiv.org/pdf/1911.11907.pdf)) - - [GhostNet_x0_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x0_5_pretrained.pdparams) - - [GhostNet_x1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_0_pretrained.pdparams) - - [GhostNet_x1_3](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_pretrained.pdparams) - - [GhostNet_x1_3_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_ssld_pretrained.pdparams) - - MixNet 系列[[29](#ref29)]([论文地址](https://arxiv.org/pdf/1907.09595.pdf)) - - [MixNet_S](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_S_pretrained.pdparams) - - [MixNet_M](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_M_pretrained.pdparams) - - [MixNet_L](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_L_pretrained.pdparams) - - ReXNet 系列[[30](#ref30)]([论文地址](https://arxiv.org/pdf/2007.00992.pdf)) - - [ReXNet_1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_0_pretrained.pdparams) - - [ReXNet_1_3](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_3_pretrained.pdparams) - - [ReXNet_1_5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_5_pretrained.pdparams) - - [ReXNet_2_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_2_0_pretrained.pdparams) - - [ReXNet_3_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_3_0_pretrained.pdparams) - - -- SEResNeXt 与 Res2Net 系列 - - ResNeXt 系列[[7](#ref7)]([论文地址](https://arxiv.org/abs/1611.05431)) - - [ResNeXt50_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_32x4d_pretrained.pdparams) - - [ResNeXt50_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_64x4d_pretrained.pdparams) - - [ResNeXt101_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x4d_pretrained.pdparams) - - [ResNeXt101_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_64x4d_pretrained.pdparams) - - [ResNeXt152_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_32x4d_pretrained.pdparams) - - [ResNeXt152_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_64x4d_pretrained.pdparams) - - ResNeXt_vd 系列 - - [ResNeXt50_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_32x4d_pretrained.pdparams) - - [ResNeXt50_vd_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_64x4d_pretrained.pdparams) - - [ResNeXt101_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_32x4d_pretrained.pdparams) - - [ResNeXt101_vd_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_64x4d_pretrained.pdparams) - - [ResNeXt152_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_32x4d_pretrained.pdparams) - - [ResNeXt152_vd_64x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_64x4d_pretrained.pdparams) - - SE_ResNet_vd 系列[[8](#ref8)]([论文地址](https://arxiv.org/abs/1709.01507)) - - [SE_ResNet18_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet18_vd_pretrained.pdparams) - - [SE_ResNet34_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet34_vd_pretrained.pdparams) - - [SE_ResNet50_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet50_vd_pretrained.pdparams) - - SE_ResNeXt 系列 - - [SE_ResNeXt50_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_32x4d_pretrained.pdparams) - - [SE_ResNeXt101_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt101_32x4d_pretrained.pdparams) - - SE_ResNeXt_vd 系列 - - [SE_ResNeXt50_vd_32x4d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_vd_32x4d_pretrained.pdparams) - - [SENet154_vd](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SENet154_vd_pretrained.pdparams) - - Res2Net 系列[[9](#ref9)]([论文地址](https://arxiv.org/abs/1904.01169)) - - [Res2Net50_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_26w_4s_pretrained.pdparams) - - [Res2Net50_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_pretrained.pdparams) - - [Res2Net50_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_ssld_pretrained.pdparams) - - [Res2Net50_14w_8s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_14w_8s_pretrained.pdparams) - - [Res2Net101_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net101_vd_26w_4s_pretrained.pdparams) - - [Res2Net101_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net101_vd_26w_4s_ssld_pretrained.pdparams) - - [Res2Net200_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net200_vd_26w_4s_pretrained.pdparams) - - [Res2Net200_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net200_vd_26w_4s_ssld_pretrained.pdparams) - - -- Inception 系列 - - GoogLeNet 系列[[10](#ref10)]([论文地址](https://arxiv.org/pdf/1409.4842.pdf)) - - [GoogLeNet](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GoogLeNet_pretrained.pdparams) - - InceptionV3 系列[[26](#ref26)]([论文地址](https://arxiv.org/abs/1512.00567)) - - [InceptionV3](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV3_pretrained.pdparams) - - InceptionV4 系列[[11](#ref11)]([论文地址](https://arxiv.org/abs/1602.07261)) - - [InceptionV4](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV4_pretrained.pdparams) - - Xception 系列[[12](#ref12)]([论文地址](http://openaccess.thecvf.com/content_cvpr_2017/html/Chollet_Xception_Deep_Learning_CVPR_2017_paper.html)) - - [Xception41](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_pretrained.pdparams) - - [Xception41_deeplab](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_deeplab_pretrained.pdparams) - - [Xception65](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_pretrained.pdparams) - - [Xception65_deeplab](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_deeplab_pretrained.pdparams) - - [Xception71](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception71_pretrained.pdparams) - - -- HRNet 系列 - - HRNet 系列[[13](#ref13)]([论文地址](https://arxiv.org/abs/1908.07919)) - - [HRNet_W18_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W18_C_pretrained.pdparams) - - [HRNet_W18_C_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W18_C_ssld_pretrained.pdparams) - - [HRNet_W30_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W30_C_pretrained.pdparams) - - [HRNet_W32_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W32_C_pretrained.pdparams) - - [HRNet_W40_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W40_C_pretrained.pdparams) - - [HRNet_W44_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W44_C_pretrained.pdparams) - - [HRNet_W48_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W48_C_pretrained.pdparams) - - [HRNet_W48_C_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W48_C_ssld_pretrained.pdparams) - - [HRNet_W64_C](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HRNet_W64_C_pretrained.pdparams) - - [SE_HRNet_W64_C_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_HRNet_W64_C_ssld_pretrained.pdparams) - -- DPN 与 DenseNet 系列 - - DPN 系列[[14](#ref14)]([论文地址](https://arxiv.org/abs/1707.01629)) - - [DPN68](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN68_pretrained.pdparams) - - [DPN92](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN92_pretrained.pdparams) - - [DPN98](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN98_pretrained.pdparams) - - [DPN107](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN107_pretrained.pdparams) - - [DPN131](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN131_pretrained.pdparams) - - DenseNet 系列[[15](#ref15)]([论文地址](https://arxiv.org/abs/1608.06993)) - - [DenseNet121](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet121_pretrained.pdparams) - - [DenseNet161](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet161_pretrained.pdparams) - - [DenseNet169](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet169_pretrained.pdparams) - - [DenseNet201](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet201_pretrained.pdparams) - - [DenseNet264](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet264_pretrained.pdparams) - - -- EfficientNet 与 ResNeXt101_wsl 系列 - - EfficientNet 系列[[16](#ref16)]([论文地址](https://arxiv.org/abs/1905.11946)) - - [EfficientNetB0_small](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_small_pretrained.pdparams) - - [EfficientNetB0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_pretrained.pdparams) - - [EfficientNetB1](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB1_pretrained.pdparams) - - [EfficientNetB2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB2_pretrained.pdparams) - - [EfficientNetB3](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB3_pretrained.pdparams) - - [EfficientNetB4](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB4_pretrained.pdparams) - - [EfficientNetB5](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB5_pretrained.pdparams) - - [EfficientNetB6](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB6_pretrained.pdparams) - - [EfficientNetB7](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB7_pretrained.pdparams) - - ResNeXt101_wsl 系列[[17](#ref17)]([论文地址](https://arxiv.org/abs/1805.00932)) - - [ResNeXt101_32x8d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x8d_wsl_pretrained.pdparams) - - [ResNeXt101_32x16d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x16d_wsl_pretrained.pdparams) - - [ResNeXt101_32x32d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x32d_wsl_pretrained.pdparams) - - [ResNeXt101_32x48d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x48d_wsl_pretrained.pdparams) - - [Fix_ResNeXt101_32x48d_wsl](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Fix_ResNeXt101_32x48d_wsl_pretrained.pdparams) - -- ResNeSt 与 RegNet 系列 - - ResNeSt 系列[[24](#ref24)]([论文地址](https://arxiv.org/abs/2004.08955)) - - [ResNeSt50_fast_1s1x64d](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_fast_1s1x64d_pretrained.pdparams) - - [ResNeSt50](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_pretrained.pdparams) - - RegNet 系列[[25](#ref25)]([paper link](https://arxiv.org/abs/2003.13678)) - - [RegNetX_4GF](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RegNetX_4GF_pretrained.pdparams) - -- Transformer 系列 - - Swin-transformer 系列[[27](#ref27)]([论文地址](https://arxiv.org/pdf/2103.14030.pdf)) - - [SwinTransformer_tiny_patch4_window7_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_tiny_patch4_window7_224_pretrained.pdparams) - - [SwinTransformer_small_patch4_window7_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_small_patch4_window7_224_pretrained.pdparams) - - [SwinTransformer_base_patch4_window7_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_pretrained.pdparams) - - [SwinTransformer_base_patch4_window12_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_pretrained.pdparams) - - [SwinTransformer_base_patch4_window7_224_22k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_22k_pretrained.pdparams) - - [SwinTransformer_base_patch4_window7_224_22kto1k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_22kto1k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window12_384_22k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window12_384_22kto1k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window7_224_22k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22k_pretrained.pdparams) - - [SwinTransformer_large_patch4_window7_224_22kto1k](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams) - - ViT 系列[[31](#ref31)]([论文地址](https://arxiv.org/pdf/2010.11929.pdf)) - - [ViT_small_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_small_patch16_224_pretrained.pdparams) - - [ViT_base_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_224_pretrained.pdparams) - - [ViT_base_patch16_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_384_pretrained.pdparams) - - [ViT_base_patch32_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch32_384_pretrained.pdparams) - - [ViT_large_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_224_pretrained.pdparams) - - [ViT_large_patch16_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_384_pretrained.pdparams) - - [ViT_large_patch32_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch32_384_pretrained.pdparams) - - DeiT 系列[[32](#ref32)]([论文地址](https://arxiv.org/pdf/2012.12877.pdf)) - - [DeiT_tiny_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_patch16_224_pretrained.pdparams) - - [DeiT_small_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_small_patch16_224_pretrained.pdparams) - - [DeiT_base_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_224_pretrained.pdparams) - - [DeiT_base_patch16_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_384_pretrained.pdparams) - - [DeiT_tiny_distilled_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_distilled_patch16_224_pretrained.pdparams) - - [DeiT_small_distilled_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_small_distilled_patch16_224_pretrained.pdparams) - - [DeiT_base_distilled_patch16_224](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_distilled_patch16_224_pretrained.pdparams) - - [DeiT_base_distilled_patch16_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_distilled_patch16_384_pretrained.pdparams) - - LeViT 系列[[33](#ref33)]([论文地址](https://arxiv.org/pdf/2104.01136.pdf)) - - [LeViT_128S](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams) - - [LeViT_128](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128_pretrained.pdparams) - - [LeViT_192](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_192_pretrained.pdparams) - - [LeViT_256](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_256_pretrained.pdparams) - - [LeViT_384](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_384_pretrained.pdparams) - - Twins 系列[[34](#ref34)]([论文地址](https://arxiv.org/pdf/2104.13840.pdf)) - - [pcpvt_small](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_small_pretrained.pdparams) - - [pcpvt_base](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_base_pretrained.pdparams) - - [pcpvt_large](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_large_pretrained.pdparams) - - [alt_gvt_small](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_small_pretrained.pdparams) - - [alt_gvt_base](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_base_pretrained.pdparams) - - [alt_gvt_large](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_large_pretrained.pdparams) - - TNT 系列[[35](#ref35)]([论文地址](https://arxiv.org/pdf/2103.00112.pdf)) - - [TNT_small](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/TNT_small_pretrained.pdparams) - -- 其他模型 - - AlexNet 系列[[18](#ref18)]([论文地址](https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf)) - - [AlexNet](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/AlexNet_pretrained.pdparams) - - SqueezeNet 系列[[19](#ref19)]([论文地址](https://arxiv.org/abs/1602.07360)) - - [SqueezeNet1_0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_0_pretrained.pdparams) - - [SqueezeNet1_1](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_1_pretrained.pdparams) - - VGG 系列[[20](#ref20)]([论文地址](https://arxiv.org/abs/1409.1556)) - - [VGG11](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG11_pretrained.pdparams) - - [VGG13](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG13_pretrained.pdparams) - - [VGG16](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG16_pretrained.pdparams) - - [VGG19](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/VGG19_pretrained.pdparams) - - DarkNet 系列[[21](#ref21)]([论文地址](https://arxiv.org/abs/1506.02640)) - - [DarkNet53](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DarkNet53_pretrained.pdparams) - - RepVGG 系列[[36](#ref36)]([论文地址](https://arxiv.org/pdf/2101.03697.pdf)) - - [RepVGG_A0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A0_pretrained.pdparams) - - [RepVGG_A1](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A1_pretrained.pdparams) - - [RepVGG_A2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A2_pretrained.pdparams) - - [RepVGG_B0](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B0_pretrained.pdparams) - - [RepVGG_B1s](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1_pretrained.pdparams) - - [RepVGG_B2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2_pretrained.pdparams) - - [RepVGG_B1g2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g2_pretrained.pdparams) - - [RepVGG_B1g4](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g4_pretrained.pdparams) - - [RepVGG_B2g4](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2g4_pretrained.pdparams) - - [RepVGG_B3g4](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3g4_pretrained.pdparams) - - HarDNet 系列[[37](#ref37)]([论文地址](https://arxiv.org/pdf/1909.00948.pdf)) - - [HarDNet39_ds](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet39_ds_pretrained.pdparams) - - [HarDNet68_ds](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_ds_pretrained.pdparams) - - [HarDNet68](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_pretrained.pdparams) - - [HarDNet85](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet85_pretrained.pdparams) - - DLA 系列[[38](#ref38)]([论文地址](https://arxiv.org/pdf/1707.06484.pdf)) - - [DLA102](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102_pretrained.pdparams) - - [DLA102x2](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x2_pretrained.pdparams) - - [DLA102x](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x_pretrained.pdparams) - - [DLA169](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA169_pretrained.pdparams) - - [DLA34](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA34_pretrained.pdparams) - - [DLA46_c](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA46_c_pretrained.pdparams) - - [DLA60](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60_pretrained.pdparams) - - [DLA60x_c](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_c_pretrained.pdparams) - - [DLA60x](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_pretrained.pdparams) - - RedNet 系列[[39](#ref39)]([论文地址](https://arxiv.org/pdf/2103.06255.pdf)) - - [RedNet26](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet26_pretrained.pdparams) - - [RedNet38](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet38_pretrained.pdparams) - - [RedNet50](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet50_pretrained.pdparams) - - [RedNet101](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet101_pretrained.pdparams) - - [RedNet152](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet152_pretrained.pdparams) - - - -**注意**：以上模型中 EfficientNetB1-B7 的预训练模型转自[pytorch 版 EfficientNet](https://github.com/lukemelas/EfficientNet-PyTorch)，ResNeXt101_wsl 系列预训练模型转自[官方 repo](https://github.com/facebookresearch/WSL-Images)，剩余预训练模型均基于飞桨训练得到的，并在 configs 里给出了相应的训练超参数。 - - - -## 4. 参考文献 - - -[1] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778. - -[2] He T, Zhang Z, Zhang H, et al. Bag of tricks for image classification with convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2019: 558-567. - -[3] Howard A, Sandler M, Chu G, et al. Searching for mobilenetv3[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1314-1324. - -[4] Sandler M, Howard A, Zhu M, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 4510-4520. - -[5] Howard A G, Zhu M, Chen B, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[J]. arXiv preprint arXiv:1704.04861, 2017. - -[6] Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 116-131. - -[7] Xie S, Girshick R, Dollár P, et al. Aggregated residual transformations for deep neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1492-1500. - - -[8] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141. - - -[9] Gao S, Cheng M M, Zhao K, et al. Res2net: A new multi-scale backbone architecture[J]. IEEE transactions on pattern analysis and machine intelligence, 2019. - -[10] Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 1-9. - - -[11] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, inception-resnet and the impact of residual connections on learning[C]//Thirty-first AAAI conference on artificial intelligence. 2017. - -[12] Chollet F. Xception: Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1251-1258. - -[13] Wang J, Sun K, Cheng T, et al. Deep high-resolution representation learning for visual recognition[J]. arXiv preprint arXiv:1908.07919, 2019. - -[14] Chen Y, Li J, Xiao H, et al. Dual path networks[C]//Advances in neural information processing systems. 2017: 4467-4475. - -[15] Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 4700-4708. - - -[16] Tan M, Le Q V. Efficientnet: Rethinking model scaling for convolutional neural networks[J]. arXiv preprint arXiv:1905.11946, 2019. - -[17] Mahajan D, Girshick R, Ramanathan V, et al. Exploring the limits of weakly supervised pretraining[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 181-196. - -[18] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]//Advances in neural information processing systems. 2012: 1097-1105. - -[19] Iandola F N, Han S, Moskewicz M W, et al. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size[J]. arXiv preprint arXiv:1602.07360, 2016. - -[20] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014. - -[21] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788. - -[22] Ding X, Guo Y, Ding G, et al. Acnet: Strengthening the kernel skeletons for powerful cnn via asymmetric convolution blocks[C]//Proceedings of the IEEE International Conference on Computer Vision. 2019: 1911-1920. - -[23] Han K, Wang Y, Tian Q, et al. GhostNet: More features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 1580-1589. - -[24] Zhang H, Wu C, Zhang Z, et al. Resnest: Split-attention networks[J]. arXiv preprint arXiv:2004.08955, 2020. - -[25] Radosavovic I, Kosaraju R P, Girshick R, et al. Designing network design spaces[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 10428-10436. - -[26] C.Szegedy, V.Vanhoucke, S.Ioffe, J.Shlens, and Z.Wojna. Rethinking the inception architecture for computer vision. arXiv preprint arXiv:1512.00567, 2015. - -[27] Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin and Baining Guo. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. - -[28]Cheng Cui, Tingquan Gao, Shengyu Wei, Yuning Du, Ruoyu Guo, Shuilong Dong, Bin Lu, Ying Zhou, Xueying Lv, Qiwen Liu, Xiaoguang Hu, Dianhai Yu, Yanjun Ma. PP-LCNet: A Lightweight CPU Convolutional Neural Network. - -[29]Mingxing Tan, Quoc V. Le. MixConv: Mixed Depthwise Convolutional Kernels. - -[30]Dongyoon Han, Sangdoo Yun, Byeongho Heo, YoungJoon Yoo. Rethinking Channel Dimensions for Efficient Model Design. - -[31]Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby. AN IMAGE IS WORTH 16X16 WORDS: -TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE. - -[32]Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Herve Jegou. Training data-efficient image transformers & distillation through attention. - -[33]Benjamin Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Herve Jegou, Matthijs Douze. LeViT: a Vision Transformer in ConvNet’s Clothing for Faster Inference. - -[34]Xiangxiang Chu, Zhi Tian, Yuqing Wang, Bo Zhang, Haibing Ren, Xiaolin Wei, Huaxia Xia, Chunhua Shen. Twins: Revisiting the Design of Spatial Attention in Vision Transformers. - -[35]Kai Han, An Xiao, Enhua Wu, Jianyuan Guo, Chunjing Xu, Yunhe Wang. Transformer in Transformer. - -[36]Xiaohan Ding, Xiangyu Zhang, Ningning Ma, Jungong Han, Guiguang Ding, Jian Sun. RepVGG: Making VGG-style ConvNets Great Again. - -[37]Ping Chao, Chao-Yang Kao, Yu-Shan Ruan, Chien-Hsiang Huang, Youn-Long Lin. HarDNet: A Low Memory Traffic Network. - -[38]Fisher Yu, Dequan Wang, Evan Shelhamer, Trevor Darrell. Deep Layer Aggregation. - -[39]Duo Lim Jie Hu, Changhu Wang, Xiangtai Li, Qi She, Lei Zhu, Tong Zhang, Qifeng Chen. Involution: Inverting the Inherence of Convolution for Visual Recognition. diff --git a/docs/zh_CN/others/feature_visiualization.md b/docs/zh_CN/others/feature_visiualization.md index 4406e1afe5d4f52aceb1cb84f67a19204a395ef3..ee2eadcbed5d5531bcda4bbc4eef6e4229d36eac 100644 --- a/docs/zh_CN/others/feature_visiualization.md +++ b/docs/zh_CN/others/feature_visiualization.md @@ -23,7 +23,7 @@ wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_pretrained.pdparams ``` -其他模型网络结构代码及预训练模型请自行下载：[模型库](../../../ppcls/arch/backbone/)，[预训练模型](../models/models_intro.md)。 +其他模型网络结构代码及预训练模型请自行下载：[模型库](../../../ppcls/arch/backbone/)，[预训练模型](../algorithm_introduction/ImageNet_models.md)。 diff --git a/docs/zh_CN/others/update_history.md b/docs/zh_CN/others/update_history.md index 55ad1a2d290701a4ad858d642799776fd05b78f4..5ea649e52a9d53eb3aab5b8b9322d1a87920fefa 100644 --- a/docs/zh_CN/others/update_history.md +++ b/docs/zh_CN/others/update_history.md @@ -3,12 +3,8 @@ - 2021.11.1 发布[PP-ShiTu技术报告](https://arxiv.org/pdf/2111.00775.pdf)，新增饮料识别demo。 - 2021.10.23 发布轻量级图像识别系统PP-ShiTu，CPU上0.2s即可完成在10w+库的图像识别。[点击这里](../quick_start/quick_start_recognition.md)立即体验。 - 2021.09.17 发布PP-LCNet系列超轻量骨干网络模型, 在Intel CPU上，单张图像预测速度约5ms，ImageNet-1K数据集上Top1识别准确率达到80.82%，超越ResNet152的模型效果。PP-LCNet的介绍可以参考[论文](https://arxiv.org/pdf/2109.15099.pdf), 或者[PP-LCNet模型介绍](../models/PP-LCNet.md)，相关指标和预训练权重可以从 [这里](../algorithm_introduction/ImageNet_models.md)下载。 -- 2021.08.11 更新7个[FAQ](../faq_series/faq_2021_s2.md)。 -- 2021.06.29 添加Swin-transformer系列模型，ImageNet1k数据集上Top1 acc最高精度可达87.2%；支持训练预测评估与whl包部署，预训练模型可以从[这里](../models/models_intro.md)下载。 -- 2021.06.22,23,24 PaddleClas官方研发团队带来技术深入解读三日直播课。课程回放：[https://aistudio.baidu.com/aistudio/course/introduce/24519](https://aistudio.baidu.com/aistudio/course/introduce/24519) -- 2021.06.16 PaddleClas v2.2版本升级，集成Metric learning，向量检索等组件。新增商品识别、动漫人物识别、车辆识别和logo识别等4个图像识别应用。新增LeViT、Twins、TNT、DLA、HarDNet、RedNet系列30个预训练模型。 - 2021.08.11 更新 7 个[FAQ](../faq_series/faq_2021_s2.md)。 -- 2021.06.29 添加 Swin-transformer 系列模型，ImageNet1k 数据集上 Top1 acc 最高精度可达 87.2%；支持训练预测评估与 whl 包部署，预训练模型可以从[这里](../models/models_intro.md)下载。 +- 2021.06.29 添加 Swin-transformer 系列模型，ImageNet1k 数据集上 Top1 acc 最高精度可达 87.2%；支持训练预测评估与 whl 包部署，预训练模型可以从[这里](../algorithm_introduction/ImageNet_models.md)下载。 - 2021.06.22,23,24 PaddleClas 官方研发团队带来技术深入解读三日直播课。课程回放：[https://aistudio.baidu.com/aistudio/course/introduce/24519](https://aistudio.baidu.com/aistudio/course/introduce/24519) - 2021.06.16 PaddleClas v2.2 版本升级，集成 Metric learning，向量检索等组件。新增商品识别、动漫人物识别、车辆识别和 logo 识别等 4 个图像识别应用。新增 LeViT、Twins、TNT、DLA、HarDNet、RedNet 系列 30 个预训练模型。 - 2021.04.15 diff --git a/docs/zh_CN/quick_start/quick_start_classification_professional.md b/docs/zh_CN/quick_start/quick_start_classification_professional.md index 4564b059f54034db346c9ae03d4bc2cbb8a8a1e4..9a44d939ce999e291120c006801b374d302ab8ba 100644 --- a/docs/zh_CN/quick_start/quick_start_classification_professional.md +++ b/docs/zh_CN/quick_start/quick_start_classification_professional.md @@ -75,9 +75,22 @@ python3 -m paddle.distributed.launch \ 验证集的最高准确率为 0.415 左右。 -* **注意** +此处使用了多个 GPU 训练，如果只使用一个 GPU，请将 `CUDA_VISIBLE_DEVICES` 设置指定 GPU，`--gpus`设置指定 GPU，下同。例如，只使用 0 号 GPU 训练： + +```shell +export CUDA_VISIBLE_DEVICES=0 +python3 -m paddle.distributed.launch \ + --gpus="0" \ + tools/train.py \ + -c ./ppcls/configs/quick_start/professional/ResNet50_vd_CIFAR100.yaml \ + -o Global.output_dir="output_CIFAR" \ + -o Optimizer.lr.learning_rate=0.01 +``` + +* **注意**: - * 如果 GPU 卡数不是 4，验证集的准确率可能与 0.415 有差异，若需保持相当的准确率，需要将配置文件中的学习率改为`当前学习率 / 4 \* 当前卡数`。下同。 +* `--gpus`中指定的 GPU 可以是 `CUDA_VISIBLE_DEVICES` 指定的 GPU 的子集。 +* 由于初始学习率和 batch-size 需要保持线性关系，所以训练从 4 个 GPU 切换到 1 个 GPU 训练时，总 batch-size 缩减为原来的 1/4，学习率也需要缩减为原来的 1/4，所以改变了默认的学习率从 0.04 到 0.01。 @@ -157,7 +170,7 @@ python3 -m paddle.distributed.launch \ * **注意** * 其他数据增广的配置文件可以参考 `ppcls/configs/ImageNet/DataAugment/` 中的配置文件。 - * 训练 CIFAR100 的迭代轮数较少，因此进行训练时，验证集的精度指标可能会有 1% 左右的波动。 +* 训练 CIFAR100 的迭代轮数较少，因此进行训练时，验证集的精度指标可能会有 1% 左右的波动。 diff --git a/docs/zh_CN/samples/.gitkeep b/docs/zh_CN/samples/.gitkeep new file mode 100644 index 0000000000000000000000000000000000000000..8b137891791fe96927ad78e64b0aad7bded08bdc --- /dev/null +++ b/docs/zh_CN/samples/.gitkeep @@ -0,0 +1 @@ + diff --git a/docs/zh_CN/samples/Electromobile_In_Elevator_Detection/README.md b/docs/zh_CN/samples/Electromobile_In_Elevator_Detection/README.md new file mode 100644 index 0000000000000000000000000000000000000000..070136c120f44d366c7832f333d3fa0e33d75ae0 --- /dev/null +++ b/docs/zh_CN/samples/Electromobile_In_Elevator_Detection/README.md @@ -0,0 +1,7 @@ +# 电梯内电瓶车入室检测 + +近年来，电瓶车进楼入户发生的火灾事故屡见不鲜，针对该现象推出了相应的电瓶车入室检测模型，旨在从源头减少这一情况的发生。 针对室内摩托车模型可能会发生的误报情况，采用了额外的图像检索方式实现更为精确的识别。 本案例使用了飞桨图像分类开发套件中的通用图像识别系统PP-ShiTu。 + +![result](./imgs/result.png) + +注:AI Studio在线运行代码请参考[电梯内电瓶车检测全流程](https://aistudio.baidu.com/aistudio/projectdetail/3497217?contributionType=1) \ No newline at end of file diff --git a/docs/zh_CN/samples/Electromobile_In_Elevator_Detection/imgs/result.png b/docs/zh_CN/samples/Electromobile_In_Elevator_Detection/imgs/result.png new file mode 100644 index 0000000000000000000000000000000000000000..5111566619208958870411a693b331bb24a36793 Binary files /dev/null and b/docs/zh_CN/samples/Electromobile_In_Elevator_Detection/imgs/result.png differ diff --git a/docs/zh_CN/samples/Goods_Recognition/README.md b/docs/zh_CN/samples/Goods_Recognition/README.md new file mode 100644 index 0000000000000000000000000000000000000000..2e6e3cf49ba4e794ff9ff1af31c7e615c67434ca --- /dev/null +++ b/docs/zh_CN/samples/Goods_Recognition/README.md @@ -0,0 +1,7 @@ +**商品识别**，即在智能零售场景下精准快速的识别商品类别、商品属性等。当下零售市场存在降低人力及运营成本、实现24小时不间断营业等需求，商品识别可以帮助越来越多的零售门店实现智慧零售数字化转型。在智慧零售概念火爆的今天，商品识别具有非常广阔的应用场景，如**货架陈列分析**、**智能结算**、**仓库管理**、**以图搜图**等。 + +本案例使用了飞桨图像分类开发套件PaddleClas中的通用图像识别系统PP-ShiTu进行**商品识别**的实现。 + +![result](./imgs/result.jpg) + +**注**: AI Studio在线运行代码请参考[智慧商超商品识别系统](https://aistudio.baidu.com/aistudio/projectdetail/3460304) \ No newline at end of file diff --git a/docs/zh_CN/samples/Goods_Recognition/imgs/result.jpg b/docs/zh_CN/samples/Goods_Recognition/imgs/result.jpg new file mode 100644 index 0000000000000000000000000000000000000000..86e8297c5d7e2962071a888314506ef1cef27c11 Binary files /dev/null and b/docs/zh_CN/samples/Goods_Recognition/imgs/result.jpg differ diff --git a/ppcls/arch/__init__.py b/ppcls/arch/__init__.py index 2d5e29db865118a145924729e81ff423de5f174b..da21e101a27eb0db2c05b658346148bda3139c80 100644 --- a/ppcls/arch/__init__.py +++ b/ppcls/arch/__init__.py @@ -27,8 +27,9 @@ from ppcls.arch.backbone.base.theseus_layer import TheseusLayer from ppcls.utils import logger from ppcls.utils.save_load import load_dygraph_pretrain from ppcls.arch.slim import prune_model, quantize_model +from ppcls.arch.distill.afd_attention import LinearTransformStudent, LinearTransformTeacher -__all__ = ["build_model", "RecModel", "DistillationModel"] +__all__ = ["build_model", "RecModel", "DistillationModel", "AttentionModel"] def build_model(config): @@ -132,3 +133,24 @@ class DistillationModel(nn.Layer): else: result_dict[model_name] = self.model_list[idx](x, label) return result_dict + + +class AttentionModel(DistillationModel): + def __init__(self, + models=None, + pretrained_list=None, + freeze_params_list=None, + **kargs): + super().__init__(models, pretrained_list, freeze_params_list, **kargs) + + def forward(self, x, label=None): + result_dict = dict() + out = x + for idx, model_name in enumerate(self.model_name_list): + if label is None: + out = self.model_list[idx](out) + result_dict.update(out) + else: + out = self.model_list[idx](out, label) + result_dict.update(out) + return result_dict diff --git a/ppcls/arch/backbone/__init__.py b/ppcls/arch/backbone/__init__.py index dadac59a38c6b3f88de51791ffd1dc6a1e281ecf..b62b5a64df348e257beee174eeb5bff1007f1d3e 100644 --- a/ppcls/arch/backbone/__init__.py +++ b/ppcls/arch/backbone/__init__.py @@ -51,6 +51,7 @@ from ppcls.arch.backbone.model_zoo.regnet import RegNetX_200MF, RegNetX_4GF, Reg from ppcls.arch.backbone.model_zoo.vision_transformer import ViT_small_patch16_224, ViT_base_patch16_224, ViT_base_patch16_384, ViT_base_patch32_384, ViT_large_patch16_224, ViT_large_patch16_384, ViT_large_patch32_384 from ppcls.arch.backbone.model_zoo.distilled_vision_transformer import DeiT_tiny_patch16_224, DeiT_small_patch16_224, DeiT_base_patch16_224, DeiT_tiny_distilled_patch16_224, DeiT_small_distilled_patch16_224, DeiT_base_distilled_patch16_224, DeiT_base_patch16_384, DeiT_base_distilled_patch16_384 from ppcls.arch.backbone.model_zoo.swin_transformer import SwinTransformer_tiny_patch4_window7_224, SwinTransformer_small_patch4_window7_224, SwinTransformer_base_patch4_window7_224, SwinTransformer_base_patch4_window12_384, SwinTransformer_large_patch4_window7_224, SwinTransformer_large_patch4_window12_384 +from ppcls.arch.backbone.model_zoo.cswin_transformer import CSWinTransformer_tiny_224, CSWinTransformer_small_224, CSWinTransformer_base_224, CSWinTransformer_large_224, CSWinTransformer_base_384, CSWinTransformer_large_384 from ppcls.arch.backbone.model_zoo.mixnet import MixNet_S, MixNet_M, MixNet_L from ppcls.arch.backbone.model_zoo.rexnet import ReXNet_1_0, ReXNet_1_3, ReXNet_1_5, ReXNet_2_0, ReXNet_3_0 from ppcls.arch.backbone.model_zoo.gvt import pcpvt_small, pcpvt_base, pcpvt_large, alt_gvt_small, alt_gvt_base, alt_gvt_large @@ -61,6 +62,9 @@ from ppcls.arch.backbone.model_zoo.tnt import TNT_small from ppcls.arch.backbone.model_zoo.hardnet import HarDNet68, HarDNet85, HarDNet39_ds, HarDNet68_ds from ppcls.arch.backbone.model_zoo.cspnet import CSPDarkNet53 from ppcls.arch.backbone.model_zoo.pvt_v2 import PVT_V2_B0, PVT_V2_B1, PVT_V2_B2_Linear, PVT_V2_B2, PVT_V2_B3, PVT_V2_B4, PVT_V2_B5 +from ppcls.arch.backbone.model_zoo.mobilevit import MobileViT_XXS, MobileViT_XS, MobileViT_S +from ppcls.arch.backbone.model_zoo.repvgg import RepVGG_A0, RepVGG_A1, RepVGG_A2, RepVGG_B0, RepVGG_B1, RepVGG_B2, RepVGG_B1g2, RepVGG_B1g4, RepVGG_B2g4, RepVGG_B3g4 +from ppcls.arch.backbone.model_zoo.van import VAN_tiny from ppcls.arch.backbone.variant_models.resnet_variant import ResNet50_last_stage_stride1 from ppcls.arch.backbone.variant_models.vgg_variant import VGG19Sigmoid from ppcls.arch.backbone.variant_models.pp_lcnet_variant import PPLCNet_x2_5_Tanh diff --git a/ppcls/arch/backbone/base/theseus_layer.py b/ppcls/arch/backbone/base/theseus_layer.py index 908d94445e035394f9b7b3f1fc72a8431435b223..9f3f596a85a4ca7336779b88e77053bf19458c7d 100644 --- a/ppcls/arch/backbone/base/theseus_layer.py +++ b/ppcls/arch/backbone/base/theseus_layer.py @@ -35,7 +35,7 @@ class TheseusLayer(nn.Layer): self.quanter = None def _return_dict_hook(self, layer, input, output): - res_dict = {"output": output} + res_dict = {"logits": output} # 'list' is needed to avoid error raised by popping self.res_dict for res_key in list(self.res_dict): # clear the res_dict because the forward process may change according to input diff --git a/ppcls/arch/backbone/legendary_models/hrnet.py b/ppcls/arch/backbone/legendary_models/hrnet.py index c3f77590cfcbeaf95d68eb8f6e829f21454e4374..399131513028cf43fd4360c859829278531ccd10 100644 --- a/ppcls/arch/backbone/legendary_models/hrnet.py +++ b/ppcls/arch/backbone/legendary_models/hrnet.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1908.07919 + from __future__ import absolute_import from __future__ import division from __future__ import print_function @@ -459,6 +461,7 @@ class HRNet(TheseusLayer): self.avg_pool = nn.AdaptiveAvgPool2D(1) stdv = 1.0 / math.sqrt(2048 * 1.0) + self.flatten = nn.Flatten(start_axis=1, stop_axis=-1) self.fc = nn.Linear( 2048, @@ -496,7 +499,7 @@ class HRNet(TheseusLayer): y = self.conv_last(y) y = self.avg_pool(y) - y = paddle.reshape(y, shape=[-1, y.shape[1]]) + y = self.flatten(y) y = self.fc(y) return y diff --git a/ppcls/arch/backbone/legendary_models/inception_v3.py b/ppcls/arch/backbone/legendary_models/inception_v3.py index 5575f8c997e1488aaf25d5eb05e8e58726b708ad..74a3d1fd210aa34d5e5615b6529b9c02d0dfc410 100644 --- a/ppcls/arch/backbone/legendary_models/inception_v3.py +++ b/ppcls/arch/backbone/legendary_models/inception_v3.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1512.00567v3 + from __future__ import absolute_import, division, print_function import math import paddle diff --git a/ppcls/arch/backbone/legendary_models/mobilenet_v1.py b/ppcls/arch/backbone/legendary_models/mobilenet_v1.py index 9767d69b3132d879b6654a87c36193bf1d82fd98..3d6caeb0f049e86997d2da281208e3963e2d28a2 100644 --- a/ppcls/arch/backbone/legendary_models/mobilenet_v1.py +++ b/ppcls/arch/backbone/legendary_models/mobilenet_v1.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1704.04861 + from __future__ import absolute_import, division, print_function from paddle import ParamAttr diff --git a/ppcls/arch/backbone/legendary_models/mobilenet_v3.py b/ppcls/arch/backbone/legendary_models/mobilenet_v3.py index 836c54cd2075464113b61e8d926ecef7ec03af54..b7fc7e9f75db79338af9211782ff7a3c1525b222 100644 --- a/ppcls/arch/backbone/legendary_models/mobilenet_v3.py +++ b/ppcls/arch/backbone/legendary_models/mobilenet_v3.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1905.02244 + from __future__ import absolute_import, division, print_function import paddle diff --git a/ppcls/arch/backbone/legendary_models/resnet.py b/ppcls/arch/backbone/legendary_models/resnet.py index 74c5c5fa64de2ffbc884d445ed770b4f2a61985c..643e860faf022000453e00cad637ef1ad572e0dc 100644 --- a/ppcls/arch/backbone/legendary_models/resnet.py +++ b/ppcls/arch/backbone/legendary_models/resnet.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/pdf/1512.03385 + from __future__ import absolute_import, division, print_function import numpy as np @@ -276,6 +278,7 @@ class ResNet(TheseusLayer): config, stages_pattern, version="vb", + stem_act="relu", class_num=1000, lr_mult_list=[1.0, 1.0, 1.0, 1.0, 1.0], data_format="NCHW", @@ -309,13 +312,13 @@ class ResNet(TheseusLayer): [[input_image_channel, 32, 3, 2], [32, 32, 3, 1], [32, 64, 3, 1]] } - self.stem = nn.Sequential(* [ + self.stem = nn.Sequential(*[ ConvBNLayer( num_channels=in_c, num_filters=out_c, filter_size=k, stride=s, - act="relu", + act=stem_act, lr_mult=self.lr_mult_list[0], data_format=data_format) for in_c, out_c, k, s in self.stem_cfg[version] diff --git a/ppcls/arch/backbone/legendary_models/vgg.py b/ppcls/arch/backbone/legendary_models/vgg.py index 74d5cfad6533801efe5defcf1f3b646a84118561..b71249616456008c77818b2d0a16b1f7a6337143 100644 --- a/ppcls/arch/backbone/legendary_models/vgg.py +++ b/ppcls/arch/backbone/legendary_models/vgg.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1409.1556 + from __future__ import absolute_import, division, print_function import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/alexnet.py b/ppcls/arch/backbone/model_zoo/alexnet.py index b44901a638039b09230aab97930d00b6b501fa44..90e1d7e1dbff4f4f9331a37c3cffe3872997da07 100644 --- a/ppcls/arch/backbone/model_zoo/alexnet.py +++ b/ppcls/arch/backbone/model_zoo/alexnet.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://proceedings.neurips.cc/paper/2012/file/c399862d3b9d6b76c8436e924a68c45b-Paper.pdf + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/cspnet.py b/ppcls/arch/backbone/model_zoo/cspnet.py index ab5021fc6f1abe79103b5fff7e12faa17aafc465..bf4e061e3835d2f07c84baa5e8af97d2192a6d1e 100644 --- a/ppcls/arch/backbone/model_zoo/cspnet.py +++ b/ppcls/arch/backbone/model_zoo/cspnet.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was heavily based on https://github.com/rwightman/pytorch-image-models +# reference: https://arxiv.org/abs/1911.11929 import paddle import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/cswin_transformer.py b/ppcls/arch/backbone/model_zoo/cswin_transformer.py new file mode 100644 index 0000000000000000000000000000000000000000..429edbe1aa18bc3704781ab1cb9f1b4b31bca69f --- /dev/null +++ b/ppcls/arch/backbone/model_zoo/cswin_transformer.py @@ -0,0 +1,651 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/BR-IDL/PaddleViT/blob/develop/image_classification/CSwin/cswin.py +# reference: https://arxiv.org/abs/2107.00652 + +import copy +import numpy as np +import paddle +import paddle.nn as nn +from .vision_transformer import trunc_normal_, zeros_, ones_, to_2tuple, DropPath, Identity + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "CSWinTransformer_tiny_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_tiny_224_pretrained.pdparams", + "CSWinTransformer_small_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_small_224_pretrained.pdparams", + "CSWinTransformer_base_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_224_pretrained.pdparams", + "CSWinTransformer_large_224": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_224_pretrained.pdparams", + "CSWinTransformer_base_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_384_pretrained.pdparams", + "CSWinTransformer_large_384": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_384_pretrained.pdparams", +} + +__all__ = list(MODEL_URLS.keys()) + + +class PatchEmbedding(nn.Layer): + """CSwin Patch Embedding + This patch embedding has a 7x7 conv + layernorm, the output tensor + is reshaped to [Batch, H*W, embed_dim]. Note that the patch is applied + by a conv with overlap (using patch_stride). + Args: + patch_stride: int, patch stride size, default: 4 + in_channels: int, number of channels of input image, default: 3 + embed_dim: int, output feature dimension, default: 96 + """ + + def __init__(self, patch_stride=4, in_channels=3, embed_dim=96): + super().__init__() + self.patch_embed = nn.Conv2D( + in_channels=in_channels, + out_channels=embed_dim, + kernel_size=7, + stride=patch_stride, + padding=2) + + self.norm = nn.LayerNorm(embed_dim) + + def forward(self, x): + x = self.patch_embed( + x) # [batch, embed_dim, h, w], h = w = image_size / 4 + x = x.flatten(start_axis=2, stop_axis=-1) # [batch, embed_dim, h*w] + x = x.transpose([0, 2, 1]) # [batch, h*w, embed_dim] + x = self.norm(x) + return x + + +class Mlp(nn.Layer): + """ MLP module + Impl using nn.Linear and activation is GELU, dropout is applied. + Ops: fc -> act -> dropout -> fc -> dropout + Attributes: + fc1: nn.Linear + fc2: nn.Linear + act: GELU + dropout1: dropout after fc1 + dropout2: dropout after fc2 + """ + + def __init__(self, in_features, hidden_features, dropout): + super().__init__() + self.fc1 = nn.Linear(in_features, hidden_features) + self.fc2 = nn.Linear(hidden_features, in_features) + self.act = nn.GELU() + self.dropout = nn.Dropout(dropout) + + def forward(self, x): + x = self.fc1(x) + x = self.act(x) + x = self.dropout(x) + x = self.fc2(x) + x = self.dropout(x) + return x + + +def img2windows(img, h_split, w_split): + """Convert input tensor into split stripes + Args: + img: tensor, image tensor with shape [B, C, H, W] + h_split: int, splits width in height direction + w_split: int, splits width in width direction + Returns: + out: tensor, splitted image + """ + B, C, H, W = img.shape + out = img.reshape([B, C, H // h_split, h_split, W // w_split, w_split]) + out = out.transpose( + [0, 2, 4, 3, 5, 1]) # [B, H//h_split, W//w_split, h_split, w_split, C] + out = out.reshape([-1, h_split * w_split, + C]) # [B, H//h_split, W//w_split, h_split*w_split, C] + return out + + +def windows2img(img_splits, h_split, w_split, img_h, img_w): + """Convert splitted stripes back + Args: + img_splits: tensor, image tensor with shape [B, C, H, W] + h_split: int, splits width in height direction + w_split: int, splits width in width direction + img_h: int, original tensor height + img_w: int, original tensor width + Returns: + img: tensor, original tensor + """ + B = paddle.to_tensor(img_splits.shape[0] // + (img_h // h_split * img_w // w_split), "int32") + img = img_splits.reshape([ + B, img_h // h_split, img_w // w_split, h_split, w_split, + img_splits.shape[-1] + ]) + img = img.transpose( + [0, 1, 3, 2, 4, + 5]) #[B,img_h//h_split, h_split, img_w//w_split, w_split,C] + img = img.reshape( + [B, img_h, img_w, img_splits.shape[-1]]) # [B, img_h, img_w, C] + return img + + +class LePEAttention(nn.Layer): + """Cross Shaped Window self-attention with Locally enhanced positional encoding""" + + def __init__(self, + dim, + resolution, + h_split=7, + w_split=7, + num_heads=8, + attention_dropout=0., + dropout=0., + qk_scale=None): + super().__init__() + self.dim = dim + self.resolution = resolution + self.num_heads = num_heads + self.dim_head = dim // num_heads + self.scale = qk_scale or self.dim_head**-0.5 + self.h_split = h_split + self.w_split = w_split + + self.get_v = nn.Conv2D( + in_channels=dim, + out_channels=dim, + kernel_size=3, + stride=1, + padding=1, + groups=dim) + + self.softmax = nn.Softmax(axis=-1) + self.attn_dropout = nn.Dropout(attention_dropout) + + def im2cswin(self, x): + B, HW, C = x.shape + H = W = int(np.sqrt(HW)) + x = x.transpose([0, 2, 1]) # [B, C, H*W] + x = x.reshape([B, C, H, W]) # [B, C, H, W] + x = img2windows(x, self.h_split, self.w_split) + x = x.reshape( + [-1, self.h_split * self.w_split, self.num_heads, self.dim_head]) + x = x.transpose([0, 2, 1, 3]) + return x + + def get_lepe(self, x, func): + """Locally Enhanced Positional Encoding (LePE) + This module applies a depthwise conv on V and returns the lepe + Args: + x: tensor, the input tensor V + func: nn.Layer, a depth wise conv of kernel 3 stride 1 and padding 1 + """ + B, HW, C = x.shape + H = W = int(np.sqrt(HW)) + h_split = self.h_split + w_split = self.w_split + + x = x.transpose([0, 2, 1]) # [B, C, H*W] + x = x.reshape([B, C, H, W]) # [B, C, H, W] + x = x.reshape([B, C, H // h_split, h_split, W // w_split, w_split]) + x = x.transpose( + [0, 2, 4, 1, 3, + 5]) # [B, H//h_split, W//w_split, C, h_split, w_split] + x = x.reshape( + [-1, C, h_split, + w_split]) # [B*(H//h_split)*(W//w_split), h_split, w_split] + + lepe = func(x) # depth wise conv does not change shape + #lepe = lepe.reshape([-1, self.num_heads, C // self.num_heads, h_split * w_split]) + lepe = lepe.reshape( + [-1, self.num_heads, self.dim_head, h_split * w_split]) + lepe = lepe.transpose( + [0, 1, 3, 2]) # [B, num_heads, h_spllit*w_split, dim_head] + + x = x.reshape([-1, self.num_heads, self.dim_head, h_split * w_split]) + x = x.transpose( + [0, 1, 3, 2]) # [B, num_heads, h_split*wsplit, dim_head] + return x, lepe + + def forward(self, q, k, v): + B, HW, C = q.shape + H = W = self.resolution + q = self.im2cswin(q) + k = self.im2cswin(k) + v, lepe = self.get_lepe(v, self.get_v) + + q = q * self.scale + attn = paddle.matmul(q, k, transpose_y=True) + attn = self.softmax(attn) + attn = self.attn_dropout(attn) + + z = paddle.matmul(attn, v) + z = z + lepe + z = z.transpose([0, 2, 1, 3]) + z = z.reshape([-1, self.h_split * self.w_split, C]) + + z = windows2img(z, self.h_split, self.w_split, H, W) + z = z.reshape([B, z.shape[1] * z.shape[2], C]) + return z + + +class CSwinBlock(nn.Layer): + """CSwin Block + CSwin block contains a LePE attention modual, a linear projection, + a mlp layer, and related norms layers. In the first 3 stages, the + LePE attention moduals used 2 branches, where horizontal and + vertical split stripes are used for self attention and a concat + op is applied to combine the outputs. The last stage does not + have branche in LePE attention. + Args: + dim: int, input feature dimension + input_resolution: int, input feature spatial size. + num_heads: int, num of attention heads in current stage + split_size: int, the split size in current stage + mlp_ratio: float, mlp ratio, mlp_hidden_dim = mlp_ratio * mlp_in_dim, default: 4. + qkv_bias: bool, if set True, qkv projection will have bias, default: True + qk_scale: float, if set, replace the orig qk_scale (dim_head ** -0.5), default: None + dropout: float, dropout rate for linear projection, default: 0 + attention_dropout: float, dropout rate for attention, default: 0 + droppath: float, drop path rate, default: 0 + split_heads: bool, if True, split heads is applied (True for 1,2,3 stages), default: True + """ + + def __init__(self, + dim, + input_resolution, + num_heads, + split_size=7, + mlp_ratio=4., + qkv_bias=False, + qk_scale=None, + attention_dropout=0., + dropout=0., + droppath=0., + split_heads=True): + super().__init__() + self.dim = dim + # NOTE: here assume image_h == imgae_w + self.input_resolution = (input_resolution, input_resolution) + self.num_heads = num_heads + self.dim_head = dim // num_heads + self.mlp_ratio = mlp_ratio + self.split_size = split_size + self.norm1 = nn.LayerNorm(dim) + self.qkv = nn.Linear( + dim, dim * 3, bias_attr=None if qkv_bias else False) + self.attns = nn.LayerList() + self.split_heads = split_heads + + num_branches = 2 if split_heads else 1 + if split_heads: # first 3 stages + splits = [self.input_resolution[0], + self.split_size] # horizantal splits + else: # last stage + splits = [self.input_resolution[0], self.input_resolution[0]] + for _ in range(num_branches): + attn = LePEAttention( + dim=dim // num_branches, + resolution=input_resolution, + h_split=splits[0], + w_split=splits[1], + num_heads=num_heads // num_branches, + qk_scale=qk_scale, + attention_dropout=attention_dropout, + dropout=dropout) + self.attns.append(copy.deepcopy(attn)) + # switch splits from horizantal to vertical + # NOTE: may need to change for different H and W + splits[0], splits[1] = splits[1], splits[0] + + self.proj = nn.Linear(dim, dim) + self.drop_path = DropPath(droppath) if droppath > 0. else Identity() + + self.norm2 = nn.LayerNorm(dim) + self.mlp = Mlp(in_features=dim, + hidden_features=int(dim * mlp_ratio), + dropout=dropout) + + def chunk_qkv(self, x, chunks=1, axis=-1): + x = x.chunk(chunks, axis=axis) + return x + + def forward(self, x): + H, W = self.input_resolution + B, HW, C = x.shape + # cswin attention + h = x + x = self.norm1(x) + qkv = self.qkv(x).chunk(3, axis=-1) # qkv is a tuple of [q, k, v] + + if self.split_heads: + q, k, v = map(self.chunk_qkv, qkv, + (2, 2, 2)) # map requries list/tuple inputs + else: + q, k, v = map(lambda x: [x], qkv) + + if self.split_heads: # first 3 stages + h_attn = self.attns[0](q[0], k[0], v[0]) + w_attn = self.attns[1](q[1], k[1], v[1]) + attn = paddle.concat([h_attn, w_attn], axis=2) + else: # last stage + attn = self.attns[0](q[0], k[0], v[0]) + attn = self.proj(attn) + attn = self.drop_path(attn) + x = h + attn + # mlp + residual + h = x + x = self.norm2(x) + x = self.mlp(x) + x = self.drop_path(x) + x = h + x + return x + + +class MergeBlock(nn.Layer): + def __init__(self, dim_in, dim_out): + super().__init__() + self.conv = nn.Conv2D( + in_channels=dim_in, + out_channels=dim_out, + kernel_size=3, + stride=2, + padding=1) + self.norm = nn.LayerNorm(dim_out) + + def forward(self, x): + B, HW, C = x.shape + H = W = int(np.sqrt(HW)) + x = x.transpose([0, 2, 1]) # [B, C, HW] + x = x.reshape([B, C, H, W]) # [B, C, H, W] + x = self.conv(x) + new_shape = [x.shape[0], x.shape[1], + x.shape[2] * x.shape[3]] # [B, C', H*W] + x = x.reshape(new_shape) # [B, C', H*W] + x = x.transpose([0, 2, 1]) # [B, H*W, C'] + x = self.norm(x) + return x + + +class CSwinStage(nn.Layer): + """ CSwin Stage, each stage contains multi blocks + CSwin has 4 stages, the first 3 stages are using head split. The last + stage does not have head split. There is a merge block between each + 2 stages. + Args: + dim: int, input feature dimension + depth: int, number of blocks in current stage + num_heads: int, num of attention heads in current stage + split_size: int, the split size in current stage + mlp_ratio: float, mlp ratio, mlp_hidden_dim = mlp_ratio * mlp_in_dim, default: 4. + qkv_bias: bool, if set True, qkv projection will have bias, default: True + qk_scale: float, if set, replace the orig qk_scale (dim_head ** -0.5), default: None + dropout: float, dropout rate for linear projection, default: 0 + attention_dropout: float, dropout rate for attention, default: 0 + droppath: float, drop path rate, default: 0 + last_stage: bool, if current stage is the last stage, default: False + """ + + def __init__(self, + dim, + input_resolution, + depth, + num_heads, + split_size, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + dropout=0., + attention_dropout=0., + droppath=0., + last_stage=False): + super().__init__() + self.blocks = nn.LayerList() + for i in range(depth): + block = CSwinBlock( + dim=dim, + input_resolution=input_resolution, + num_heads=num_heads, + split_size=split_size, + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + attention_dropout=attention_dropout, + dropout=dropout, + droppath=droppath[i] + if isinstance(droppath, list) else droppath, + split_heads=not last_stage) + self.blocks.append(copy.deepcopy(block)) + # last stage does not need merge layer + self.merge = MergeBlock( + dim_in=dim, dim_out=dim * 2) if not last_stage else Identity() + + def forward(self, x): + for block in self.blocks: + x = block(x) + x = self.merge(x) + return x + + +class CSwinTransformer(nn.Layer): + """CSwin Transformer class + Args: + image_size: int, input image size, default: 224 + patch_stride: int, stride for patch embedding, default: 4 + in_channels: int, num of channels of input image, default: 3 + num_classes: int, num of classes, default: 1000 + embed_dim: int, embedding dim (patch embed out dim), default: 96 + depths: list/tuple(int), number of blocks in each stage, default: [2, 4, 32, 2] + splits: list/tuple(int), the split number in each stage, default: [1, 2, 7, 7] + num_heads: list/tuple(int), num of attention heads in each stage, default: [4, 8, 16, 32] + mlp_ratio: float, mlp ratio, mlp_hidden_dim = mlp_ratio * mlp_in_dim, default: 4. + qkv_bias: bool, if set True, qkv projection will have bias, default: True + qk_scale: float, if set, replace the orig qk_scale (dim_head ** -0.5), default: None + dropout: float, dropout rate for linear projection, default: 0 + attention_dropout: float, dropout rate for attention, default: 0 + droppath: float, drop path rate, default: 0 + """ + + def __init__(self, + image_size=224, + patch_stride=4, + in_channels=3, + class_num=1000, + embed_dim=96, + depths=[2, 4, 32, 2], + splits=[1, 2, 7, 7], + num_heads=[4, 8, 16, 32], + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + dropout=0., + attention_dropout=0., + droppath=0.): + super().__init__() + # token embedding + self.patch_embedding = PatchEmbedding( + patch_stride=patch_stride, + in_channels=in_channels, + embed_dim=embed_dim) + # drop path decay by stage + depth_decay = [ + x.item() for x in paddle.linspace(0, droppath, sum(depths)) + ] + dim = embed_dim + resolution = image_size // 4 + self.stages = nn.LayerList() + num_stages = len(depths) + # construct CSwin stages: each stage has multiple blocks + for stage_idx in range(num_stages): + stage = CSwinStage( + dim=dim, + input_resolution=resolution, + depth=depths[stage_idx], + num_heads=num_heads[stage_idx], + split_size=splits[stage_idx], + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + dropout=dropout, + attention_dropout=attention_dropout, + droppath=depth_decay[sum(depths[:stage_idx]):sum( + depths[:stage_idx + 1])], + last_stage=stage_idx == num_stages - 1) + self.stages.append(stage) + if stage_idx != num_stages - 1: + dim = dim * 2 + resolution = resolution // 2 + # last norm and classification head layers + self.norm = nn.LayerNorm(dim) + self.head = nn.Linear(dim, class_num) + + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + + def forward_features(self, x): + x = self.patch_embedding(x) + for stage in self.stages: + x = stage(x) + x = self.norm(x) + return paddle.mean(x, axis=1) + + def forward(self, x): + x = self.forward_features(x) + x = self.head(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def CSWinTransformer_tiny_224(pretrained=False, use_ssld=False, **kwargs): + model = CSwinTransformer( + image_size=224, + embed_dim=64, + depths=[1, 2, 21, 1], + splits=[1, 2, 7, 7], + num_heads=[2, 4, 8, 16], + droppath=0.2, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["CSWinTransformer_tiny_224"], + use_ssld=use_ssld) + return model + + +def CSWinTransformer_small_224(pretrained=False, use_ssld=False, **kwargs): + model = CSwinTransformer( + image_size=224, + embed_dim=64, + depths=[2, 4, 32, 2], + splits=[1, 2, 7, 7], + num_heads=[2, 4, 8, 16], + droppath=0.4, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["CSWinTransformer_small_224"], + use_ssld=use_ssld) + return model + + +def CSWinTransformer_base_224(pretrained=False, use_ssld=False, **kwargs): + model = CSwinTransformer( + image_size=224, + embed_dim=96, + depths=[2, 4, 32, 2], + splits=[1, 2, 7, 7], + num_heads=[4, 8, 16, 32], + droppath=0.5, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["CSWinTransformer_base_224"], + use_ssld=use_ssld) + return model + + +def CSWinTransformer_base_384(pretrained=False, use_ssld=False, **kwargs): + model = CSwinTransformer( + image_size=384, + embed_dim=96, + depths=[2, 4, 32, 2], + splits=[1, 2, 12, 12], + num_heads=[4, 8, 16, 32], + droppath=0.5, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["CSWinTransformer_base_384"], + use_ssld=use_ssld) + return model + + +def CSWinTransformer_large_224(pretrained=False, use_ssld=False, **kwargs): + model = CSwinTransformer( + image_size=224, + embed_dim=144, + depths=[2, 4, 32, 2], + splits=[1, 2, 7, 7], + num_heads=[6, 12, 24, 24], + droppath=0.5, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["CSWinTransformer_large_224"], + use_ssld=use_ssld) + return model + + +def CSWinTransformer_large_384(pretrained=False, use_ssld=False, **kwargs): + model = CSwinTransformer( + image_size=384, + embed_dim=144, + depths=[2, 4, 32, 2], + splits=[1, 2, 12, 12], + num_heads=[6, 12, 24, 24], + droppath=0.5, + **kwargs) + _load_pretrained( + pretrained, + model, + MODEL_URLS["CSWinTransformer_large_384"], + use_ssld=use_ssld) + return model diff --git a/ppcls/arch/backbone/model_zoo/darknet.py b/ppcls/arch/backbone/model_zoo/darknet.py index 75aafd85b1a3d08594f7ee551d982c30b54fd938..34ca9a2e71dbaee94431853a94495f639f4e53cf 100644 --- a/ppcls/arch/backbone/model_zoo/darknet.py +++ b/ppcls/arch/backbone/model_zoo/darknet.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1804.02767 + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/densenet.py b/ppcls/arch/backbone/model_zoo/densenet.py index 7e6e20251fc15e4a59b294ef9fe53a9de7f67674..d658a18f8f18faea6faece1fa1f56ffc61680f68 100644 --- a/ppcls/arch/backbone/model_zoo/densenet.py +++ b/ppcls/arch/backbone/model_zoo/densenet.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1608.06993 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/distilled_vision_transformer.py b/ppcls/arch/backbone/model_zoo/distilled_vision_transformer.py index 676a2897c87700a73f7dce0762ebf7bccbe1cdf1..e816e804204235d0bd3ab08e8e4b687787e5cd94 100644 --- a/ppcls/arch/backbone/model_zoo/distilled_vision_transformer.py +++ b/ppcls/arch/backbone/model_zoo/distilled_vision_transformer.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was heavily based on https://github.com/facebookresearch/deit +# reference: https://arxiv.org/abs/2012.12877 import paddle import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/dla.py b/ppcls/arch/backbone/model_zoo/dla.py index b1c00b2dc33970ae27911bccec6782ca8be79a47..d5ee822c480fa5c88e18dca2288c2f5b5cff2f97 100644 --- a/ppcls/arch/backbone/model_zoo/dla.py +++ b/ppcls/arch/backbone/model_zoo/dla.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/ucbdrive/dla +# reference: https://arxiv.org/abs/1707.06484 import math diff --git a/ppcls/arch/backbone/model_zoo/dpn.py b/ppcls/arch/backbone/model_zoo/dpn.py index 55953ed2064beb21d5e65634db5cecb8e3c0d948..2eb2647e88d5f3ec7a319f1b68367c189ac8ee11 100644 --- a/ppcls/arch/backbone/model_zoo/dpn.py +++ b/ppcls/arch/backbone/model_zoo/dpn.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1707.01629 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/efficientnet.py b/ppcls/arch/backbone/model_zoo/efficientnet.py index bd0cffa621f2d32fcb7c3853ccd2a51c42dc328e..9ddb90fbb077622f86f0624bdd30b8a3a9297ac9 100644 --- a/ppcls/arch/backbone/model_zoo/efficientnet.py +++ b/ppcls/arch/backbone/model_zoo/efficientnet.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/lukemelas/EfficientNet-PyTorch +# reference: https://arxiv.org/abs/1905.11946 import paddle from paddle import ParamAttr diff --git a/ppcls/arch/backbone/model_zoo/ghostnet.py b/ppcls/arch/backbone/model_zoo/ghostnet.py index 4d338c1de04480e9a899a8aa4ad8d1c14812223e..5cfa56ebd5a58c0ae1cd6d2752a7e64a21538507 100644 --- a/ppcls/arch/backbone/model_zoo/ghostnet.py +++ b/ppcls/arch/backbone/model_zoo/ghostnet.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/huawei-noah/CV-Backbones/tree/master/ghostnet_pytorch +# reference: https://arxiv.org/abs/1911.11907 import math import paddle diff --git a/ppcls/arch/backbone/model_zoo/googlenet.py b/ppcls/arch/backbone/model_zoo/googlenet.py index 22528427ea3b9afa38856d632fbc08901f3c1009..1461b1ba52fb22d29d87092af995fc8783daf7b6 100644 --- a/ppcls/arch/backbone/model_zoo/googlenet.py +++ b/ppcls/arch/backbone/model_zoo/googlenet.py @@ -1,3 +1,19 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# reference: https://arxiv.org/abs/1409.4842 + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/gvt.py b/ppcls/arch/backbone/model_zoo/gvt.py index 2af7ccf43fe8ada61da861726585b4633db64d30..cf1affea80c8572605ab32e0df4cf56788411662 100644 --- a/ppcls/arch/backbone/model_zoo/gvt.py +++ b/ppcls/arch/backbone/model_zoo/gvt.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/Meituan-AutoML/Twins +# reference: https://arxiv.org/abs/2104.13840 from functools import partial diff --git a/ppcls/arch/backbone/model_zoo/hardnet.py b/ppcls/arch/backbone/model_zoo/hardnet.py index fffd3a420eee9ced34235aeff1f00374676c2d4f..ea24cdfce571d051c39dcdb419fe52beefbaf720 100644 --- a/ppcls/arch/backbone/model_zoo/hardnet.py +++ b/ppcls/arch/backbone/model_zoo/hardnet.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/PingoLH/Pytorch-HarDNet +# reference: https://arxiv.org/abs/1909.00948 import paddle import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/inception_v4.py b/ppcls/arch/backbone/model_zoo/inception_v4.py index e0460d48b70f573e78d9e3d6ee2f58103662a70d..73b3c70b7df681e9eb131419ee0fd5749d505871 100644 --- a/ppcls/arch/backbone/model_zoo/inception_v4.py +++ b/ppcls/arch/backbone/model_zoo/inception_v4.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1602.07261 + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/levit.py b/ppcls/arch/backbone/model_zoo/levit.py index 991f832bb103ef22eaec87266f1bbf210f3708d3..a5cb534f857bfdd7ed983838ae08cbc3231e011b 100644 --- a/ppcls/arch/backbone/model_zoo/levit.py +++ b/ppcls/arch/backbone/model_zoo/levit.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/facebookresearch/LeViT +# reference: https://openaccess.thecvf.com/content/ICCV2021/html/Graham_LeViT_A_Vision_Transformer_in_ConvNets_Clothing_for_Faster_Inference_ICCV_2021_paper.html import itertools import math diff --git a/ppcls/arch/backbone/model_zoo/mixnet.py b/ppcls/arch/backbone/model_zoo/mixnet.py index c2a1adb1bc82080e42e8eb735fb1c8ea7f8eb88e..38bb7c3c6a78b2474026c8c8b07a08255ec97730 100644 --- a/ppcls/arch/backbone/model_zoo/mixnet.py +++ b/ppcls/arch/backbone/model_zoo/mixnet.py @@ -11,11 +11,8 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -""" - MixNet for ImageNet-1K, implemented in Paddle. - Original paper: 'MixConv: Mixed Depthwise Convolutional Kernels,' - https://arxiv.org/abs/1907.09595. -""" + +# reference: https://arxiv.org/abs/1907.09595 import os from inspect import isfunction diff --git a/ppcls/arch/backbone/model_zoo/mobilenet_v2.py b/ppcls/arch/backbone/model_zoo/mobilenet_v2.py index b32c0250bad614fefa36ba9e4157d5d27d29c7a9..ab2a89bb884b694a895bd284b69121c0d025a09a 100644 --- a/ppcls/arch/backbone/model_zoo/mobilenet_v2.py +++ b/ppcls/arch/backbone/model_zoo/mobilenet_v2.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1801.04381 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/mobilevit.py b/ppcls/arch/backbone/model_zoo/mobilevit.py new file mode 100644 index 0000000000000000000000000000000000000000..7c0de0e620362db85085c92e2436f9a0a0a24f47 --- /dev/null +++ b/ppcls/arch/backbone/model_zoo/mobilevit.py @@ -0,0 +1,479 @@ +# copyright (c) 2022 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was based on https://github.com/BR-IDL/PaddleViT/blob/develop/image_classification/MobileViT/mobilevit.py +# and https://github.com/apple/ml-cvnets/blob/main/cvnets/models/classification/mobilevit.py +# reference: https://arxiv.org/abs/2110.02178 + +import paddle +from paddle import ParamAttr +import paddle.nn as nn +import paddle.nn.functional as F +from paddle.nn.initializer import KaimingUniform, TruncatedNormal, Constant +import math + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "MobileViT_XXS": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XXS_pretrained.pdparams", + "MobileViT_XS": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XS_pretrained.pdparams", + "MobileViT_S": + "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_S_pretrained.pdparams", +} + + +def _init_weights_linear(): + weight_attr = ParamAttr(initializer=TruncatedNormal(std=.02)) + bias_attr = ParamAttr(initializer=Constant(0.0)) + return weight_attr, bias_attr + + +def _init_weights_layernorm(): + weight_attr = ParamAttr(initializer=Constant(1.0)) + bias_attr = ParamAttr(initializer=Constant(0.0)) + return weight_attr, bias_attr + + +class ConvBnAct(nn.Layer): + def __init__(self, + in_channels, + out_channels, + kernel_size=3, + stride=1, + padding=0, + bias_attr=False, + groups=1): + super().__init__() + self.in_channels = in_channels + self.conv = nn.Conv2D( + in_channels=in_channels, + out_channels=out_channels, + kernel_size=kernel_size, + stride=stride, + padding=padding, + groups=groups, + weight_attr=ParamAttr(initializer=KaimingUniform()), + bias_attr=bias_attr) + self.norm = nn.BatchNorm2D(out_channels) + self.act = nn.Silu() + + def forward(self, inputs): + out = self.conv(inputs) + out = self.norm(out) + out = self.act(out) + return out + + +class Identity(nn.Layer): + """ Identity layer""" + + def __init__(self): + super().__init__() + + def forward(self, inputs): + return inputs + + +class Mlp(nn.Layer): + def __init__(self, embed_dim, mlp_ratio, dropout=0.1): + super().__init__() + w_attr_1, b_attr_1 = _init_weights_linear() + self.fc1 = nn.Linear( + embed_dim, + int(embed_dim * mlp_ratio), + weight_attr=w_attr_1, + bias_attr=b_attr_1) + + w_attr_2, b_attr_2 = _init_weights_linear() + self.fc2 = nn.Linear( + int(embed_dim * mlp_ratio), + embed_dim, + weight_attr=w_attr_2, + bias_attr=b_attr_2) + + self.act = nn.Silu() + self.dropout1 = nn.Dropout(dropout) + self.dropout2 = nn.Dropout(dropout) + + def forward(self, x): + x = self.fc1(x) + x = self.act(x) + x = self.dropout1(x) + x = self.fc2(x) + x = self.dropout2(x) + return x + + +class Attention(nn.Layer): + def __init__(self, + embed_dim, + num_heads, + qkv_bias=True, + dropout=0.1, + attention_dropout=0.): + super().__init__() + self.num_heads = num_heads + self.attn_head_dim = int(embed_dim / self.num_heads) + self.all_head_dim = self.attn_head_dim * self.num_heads + + w_attr_1, b_attr_1 = _init_weights_linear() + self.qkv = nn.Linear( + embed_dim, + self.all_head_dim * 3, + weight_attr=w_attr_1, + bias_attr=b_attr_1 if qkv_bias else False) + + self.scales = self.attn_head_dim**-0.5 + + w_attr_2, b_attr_2 = _init_weights_linear() + self.proj = nn.Linear( + embed_dim, embed_dim, weight_attr=w_attr_2, bias_attr=b_attr_2) + + self.attn_dropout = nn.Dropout(attention_dropout) + self.proj_dropout = nn.Dropout(dropout) + self.softmax = nn.Softmax(axis=-1) + + def transpose_multihead(self, x): + B, P, N, d = x.shape + x = x.reshape([B, P, N, self.num_heads, d // self.num_heads]) + x = x.transpose([0, 1, 3, 2, 4]) + return x + + def forward(self, x): + b_sz, n_patches, in_channels = x.shape + qkv = self.qkv(x) + qkv = qkv.reshape([ + b_sz, n_patches, 3, self.num_heads, + qkv.shape[-1] // self.num_heads // 3 + ]) + qkv = qkv.transpose([0, 3, 2, 1, 4]) + query, key, value = qkv[:, :, 0], qkv[:, :, 1], qkv[:, :, 2] + query = query * self.scales + key = key.transpose([0, 1, 3, 2]) + # QK^T + attn = paddle.matmul(query, key) + attn = self.softmax(attn) + attn = self.attn_dropout(attn) + # weighted sum + out = paddle.matmul(attn, value) + out = out.transpose([0, 2, 1, 3]).reshape( + [b_sz, n_patches, out.shape[1] * out.shape[3]]) + out = self.proj(out) + out = self.proj_dropout(out) + return out + + +class EncoderLayer(nn.Layer): + def __init__(self, + embed_dim, + num_heads=4, + qkv_bias=True, + mlp_ratio=2.0, + dropout=0.1, + attention_dropout=0., + droppath=0.): + super().__init__() + w_attr_1, b_attr_1 = _init_weights_layernorm() + w_attr_2, b_attr_2 = _init_weights_layernorm() + + self.attn_norm = nn.LayerNorm( + embed_dim, weight_attr=w_attr_1, bias_attr=b_attr_1) + self.attn = Attention(embed_dim, num_heads, qkv_bias, dropout, + attention_dropout) + self.drop_path = DropPath(droppath) if droppath > 0. else Identity() + self.mlp_norm = nn.LayerNorm( + embed_dim, weight_attr=w_attr_2, bias_attr=b_attr_2) + self.mlp = Mlp(embed_dim, mlp_ratio, dropout) + + def forward(self, x): + h = x + x = self.attn_norm(x) + x = self.attn(x) + x = self.drop_path(x) + x = h + x + h = x + x = self.mlp_norm(x) + x = self.mlp(x) + x = self.drop_path(x) + x = x + h + return x + + +class Transformer(nn.Layer): + """Transformer block for MobileViTBlock""" + + def __init__(self, + embed_dim, + num_heads, + depth, + qkv_bias=True, + mlp_ratio=2.0, + dropout=0.1, + attention_dropout=0., + droppath=0.): + super().__init__() + depth_decay = [x.item() for x in paddle.linspace(0, droppath, depth)] + + layer_list = [] + for i in range(depth): + layer_list.append( + EncoderLayer(embed_dim, num_heads, qkv_bias, mlp_ratio, + dropout, attention_dropout, droppath)) + self.layers = nn.LayerList(layer_list) + + w_attr_1, b_attr_1 = _init_weights_layernorm() + self.norm = nn.LayerNorm( + embed_dim, weight_attr=w_attr_1, bias_attr=b_attr_1, epsilon=1e-6) + + def forward(self, x): + for layer in self.layers: + x = layer(x) + out = self.norm(x) + return out + + +class MobileV2Block(nn.Layer): + """Mobilenet v2 InvertedResidual block""" + + def __init__(self, inp, oup, stride=1, expansion=4): + super().__init__() + self.stride = stride + assert stride in [1, 2] + + hidden_dim = int(round(inp * expansion)) + self.use_res_connect = self.stride == 1 and inp == oup + + layers = [] + if expansion != 1: + layers.append(ConvBnAct(inp, hidden_dim, kernel_size=1)) + + layers.extend([ + # dw + ConvBnAct( + hidden_dim, + hidden_dim, + stride=stride, + groups=hidden_dim, + padding=1), + # pw-linear + nn.Conv2D( + hidden_dim, oup, 1, 1, 0, bias_attr=False), + nn.BatchNorm2D(oup), + ]) + + self.conv = nn.Sequential(*layers) + self.out_channels = oup + + def forward(self, x): + if self.use_res_connect: + return x + self.conv(x) + return self.conv(x) + + +class MobileViTBlock(nn.Layer): + """ MobileViTBlock for MobileViT""" + + def __init__(self, + dim, + hidden_dim, + depth, + num_heads=4, + qkv_bias=True, + mlp_ratio=2.0, + dropout=0.1, + attention_dropout=0., + droppath=0.0, + patch_size=(2, 2)): + super().__init__() + self.patch_h, self.patch_w = patch_size + + # local representations + self.conv1 = ConvBnAct(dim, dim, padding=1) + self.conv2 = nn.Conv2D( + dim, hidden_dim, kernel_size=1, stride=1, bias_attr=False) + # global representations + self.transformer = Transformer( + embed_dim=hidden_dim, + num_heads=num_heads, + depth=depth, + qkv_bias=qkv_bias, + mlp_ratio=mlp_ratio, + dropout=dropout, + attention_dropout=attention_dropout, + droppath=droppath) + + # fusion + self.conv3 = ConvBnAct(hidden_dim, dim, kernel_size=1) + self.conv4 = ConvBnAct(2 * dim, dim, padding=1) + + def forward(self, x): + h = x + x = self.conv1(x) + x = self.conv2(x) + + patch_h = self.patch_h + patch_w = self.patch_w + patch_area = int(patch_w * patch_h) + _, in_channels, orig_h, orig_w = x.shape + new_h = int(math.ceil(orig_h / self.patch_h) * self.patch_h) + new_w = int(math.ceil(orig_w / self.patch_w) * self.patch_w) + interpolate = False + + if new_w != orig_w or new_h != orig_h: + x = F.interpolate(x, size=[new_h, new_w], mode="bilinear") + interpolate = True + + num_patch_w, num_patch_h = new_w // patch_w, new_h // patch_h + num_patches = num_patch_h * num_patch_w + reshaped_x = x.reshape([-1, patch_h, num_patch_w, patch_w]) + transposed_x = reshaped_x.transpose([0, 2, 1, 3]) + reshaped_x = transposed_x.reshape( + [-1, in_channels, num_patches, patch_area]) + transposed_x = reshaped_x.transpose([0, 3, 2, 1]) + + x = transposed_x.reshape([-1, num_patches, in_channels]) + x = self.transformer(x) + x = x.reshape([-1, patch_h * patch_w, num_patches, in_channels]) + + _, pixels, num_patches, channels = x.shape + x = x.transpose([0, 3, 2, 1]) + x = x.reshape([-1, num_patch_w, patch_h, patch_w]) + x = x.transpose([0, 2, 1, 3]) + x = x.reshape( + [-1, channels, num_patch_h * patch_h, num_patch_w * patch_w]) + + if interpolate: + x = F.interpolate(x, size=[orig_h, orig_w]) + x = self.conv3(x) + x = paddle.concat((h, x), axis=1) + x = self.conv4(x) + return x + + +class MobileViT(nn.Layer): + """ MobileViT + A PaddlePaddle impl of : `MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer` - + https://arxiv.org/abs/2110.02178 + """ + + def __init__(self, + in_channels=3, + dims=[16, 32, 48, 48, 48, 64, 80, 96, 384], + hidden_dims=[96, 120, 144], + mv2_expansion=4, + class_num=1000): + super().__init__() + self.conv3x3 = ConvBnAct( + in_channels, dims[0], kernel_size=3, stride=2, padding=1) + self.mv2_block_1 = MobileV2Block( + dims[0], dims[1], expansion=mv2_expansion) + self.mv2_block_2 = MobileV2Block( + dims[1], dims[2], stride=2, expansion=mv2_expansion) + self.mv2_block_3 = MobileV2Block( + dims[2], dims[3], expansion=mv2_expansion) + self.mv2_block_4 = MobileV2Block( + dims[3], dims[4], expansion=mv2_expansion) + + self.mv2_block_5 = MobileV2Block( + dims[4], dims[5], stride=2, expansion=mv2_expansion) + self.mvit_block_1 = MobileViTBlock(dims[5], hidden_dims[0], depth=2) + + self.mv2_block_6 = MobileV2Block( + dims[5], dims[6], stride=2, expansion=mv2_expansion) + self.mvit_block_2 = MobileViTBlock(dims[6], hidden_dims[1], depth=4) + + self.mv2_block_7 = MobileV2Block( + dims[6], dims[7], stride=2, expansion=mv2_expansion) + self.mvit_block_3 = MobileViTBlock(dims[7], hidden_dims[2], depth=3) + self.conv1x1 = ConvBnAct(dims[7], dims[8], kernel_size=1) + + self.pool = nn.AdaptiveAvgPool2D(1) + self.dropout = nn.Dropout(0.1) + self.linear = nn.Linear(dims[8], class_num) + + def forward(self, x): + x = self.conv3x3(x) + x = self.mv2_block_1(x) + x = self.mv2_block_2(x) + x = self.mv2_block_3(x) + x = self.mv2_block_4(x) + + x = self.mv2_block_5(x) + x = self.mvit_block_1(x) + + x = self.mv2_block_6(x) + x = self.mvit_block_2(x) + + x = self.mv2_block_7(x) + x = self.mvit_block_3(x) + x = self.conv1x1(x) + + x = self.pool(x) + x = x.reshape(x.shape[:2]) + + x = self.dropout(x) + x = self.linear(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def MobileViT_XXS(pretrained=False, use_ssld=False, **kwargs): + model = MobileViT( + in_channels=3, + dims=[16, 16, 24, 24, 24, 48, 64, 80, 320], + hidden_dims=[64, 80, 96], + mv2_expansion=2, + **kwargs) + + _load_pretrained( + pretrained, model, MODEL_URLS["MobileViT_XXS"], use_ssld=use_ssld) + return model + + +def MobileViT_XS(pretrained=False, use_ssld=False, **kwargs): + model = MobileViT( + in_channels=3, + dims=[16, 32, 48, 48, 48, 64, 80, 96, 384], + hidden_dims=[96, 120, 144], + mv2_expansion=4, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileViT_XS"], use_ssld=use_ssld) + return model + + +def MobileViT_S(pretrained=False, use_ssld=False, **kwargs): + model = MobileViT( + in_channels=3, + dims=[16, 32, 64, 64, 64, 96, 128, 160, 640], + hidden_dims=[144, 192, 240], + mv2_expansion=4, + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["MobileViT_S"], use_ssld=use_ssld) + return model diff --git a/ppcls/arch/backbone/model_zoo/pvt_v2.py b/ppcls/arch/backbone/model_zoo/pvt_v2.py index e2fdfd4ff307263c4d2020b0893f427914eb5bf6..e6b5ff27f466f12afba198c2b4994eccb87824f6 100644 --- a/ppcls/arch/backbone/model_zoo/pvt_v2.py +++ b/ppcls/arch/backbone/model_zoo/pvt_v2.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was heavily based on https://github.com/whai362/PVT +# reference: https://arxiv.org/abs/2106.13797 from functools import partial import math diff --git a/ppcls/arch/backbone/model_zoo/rednet.py b/ppcls/arch/backbone/model_zoo/rednet.py index be84da1efab88b64fde71c05c8e7dc76b2159484..eb52621c3b5d236cf6dcaa17439383a6963469f0 100644 --- a/ppcls/arch/backbone/model_zoo/rednet.py +++ b/ppcls/arch/backbone/model_zoo/rednet.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/d-li14/involution +# reference: https://arxiv.org/abs/2103.06255 import paddle import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/regnet.py b/ppcls/arch/backbone/model_zoo/regnet.py index dc381cbcc12e85152fe3a0556731061ec42188b8..a8c2942023eecf4d80ffb6eb9c66081bc628c8c4 100644 --- a/ppcls/arch/backbone/model_zoo/regnet.py +++ b/ppcls/arch/backbone/model_zoo/regnet.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/facebookresearch/pycls +# reference: https://arxiv.org/abs/1905.13214 from __future__ import absolute_import from __future__ import division diff --git a/ppcls/arch/backbone/model_zoo/repvgg.py b/ppcls/arch/backbone/model_zoo/repvgg.py index e919a59fe6e00831e6b02a4d3c8b7cc8d8b9c432..8ff662a7f88086abeee6b7f6e0260d2d3b3cd0c1 100644 --- a/ppcls/arch/backbone/model_zoo/repvgg.py +++ b/ppcls/arch/backbone/model_zoo/repvgg.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/DingXiaoH/RepVGG +# reference: https://arxiv.org/abs/2101.03697 import paddle.nn as nn import paddle @@ -33,18 +34,12 @@ MODEL_URLS = { "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1_pretrained.pdparams", "RepVGG_B2": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2_pretrained.pdparams", - "RepVGG_B3": - "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3_pretrained.pdparams", "RepVGG_B1g2": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g2_pretrained.pdparams", "RepVGG_B1g4": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g4_pretrained.pdparams", - "RepVGG_B2g2": - "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2g2_pretrained.pdparams", "RepVGG_B2g4": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2g4_pretrained.pdparams", - "RepVGG_B3g2": - "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3g2_pretrained.pdparams", "RepVGG_B3g4": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3g4_pretrained.pdparams", } @@ -92,6 +87,8 @@ class RepVGGBlock(nn.Layer): groups=1, padding_mode='zeros'): super(RepVGGBlock, self).__init__() + self.is_repped = False + self.in_channels = in_channels self.out_channels = out_channels self.kernel_size = kernel_size @@ -127,6 +124,12 @@ class RepVGGBlock(nn.Layer): groups=groups) def forward(self, inputs): + if not self.training and not self.is_repped: + self.rep() + self.is_repped = True + if self.training and self.is_repped: + self.is_repped = False + if not self.training: return self.nonlinearity(self.rbr_reparam(inputs)) @@ -137,7 +140,7 @@ class RepVGGBlock(nn.Layer): return self.nonlinearity( self.rbr_dense(inputs) + self.rbr_1x1(inputs) + id_out) - def eval(self): + def rep(self): if not hasattr(self, 'rbr_reparam'): self.rbr_reparam = nn.Conv2D( in_channels=self.in_channels, @@ -148,12 +151,9 @@ class RepVGGBlock(nn.Layer): dilation=self.dilation, groups=self.groups, padding_mode=self.padding_mode) - self.training = False kernel, bias = self.get_equivalent_kernel_bias() self.rbr_reparam.weight.set_value(kernel) self.rbr_reparam.bias.set_value(bias) - for layer in self.sublayers(): - layer.eval() def get_equivalent_kernel_bias(self): kernel3x3, bias3x3 = self._fuse_bn_tensor(self.rbr_dense) @@ -248,12 +248,6 @@ class RepVGG(nn.Layer): self.cur_layer_idx += 1 return nn.Sequential(*blocks) - def eval(self): - self.training = False - for layer in self.sublayers(): - layer.training = False - layer.eval() - def forward(self, x): out = self.stage0(x) out = self.stage1(out) @@ -367,17 +361,6 @@ def RepVGG_B2(pretrained=False, use_ssld=False, **kwargs): return model -def RepVGG_B2g2(pretrained=False, use_ssld=False, **kwargs): - model = RepVGG( - num_blocks=[4, 6, 16, 1], - width_multiplier=[2.5, 2.5, 2.5, 5], - override_groups_map=g2_map, - **kwargs) - _load_pretrained( - pretrained, model, MODEL_URLS["RepVGG_B2g2"], use_ssld=use_ssld) - return model - - def RepVGG_B2g4(pretrained=False, use_ssld=False, **kwargs): model = RepVGG( num_blocks=[4, 6, 16, 1], @@ -389,28 +372,6 @@ def RepVGG_B2g4(pretrained=False, use_ssld=False, **kwargs): return model -def RepVGG_B3(pretrained=False, use_ssld=False, **kwargs): - model = RepVGG( - num_blocks=[4, 6, 16, 1], - width_multiplier=[3, 3, 3, 5], - override_groups_map=None, - **kwargs) - _load_pretrained( - pretrained, model, MODEL_URLS["RepVGG_B3"], use_ssld=use_ssld) - return model - - -def RepVGG_B3g2(pretrained=False, use_ssld=False, **kwargs): - model = RepVGG( - num_blocks=[4, 6, 16, 1], - width_multiplier=[3, 3, 3, 5], - override_groups_map=g2_map, - **kwargs) - _load_pretrained( - pretrained, model, MODEL_URLS["RepVGG_B3g2"], use_ssld=use_ssld) - return model - - def RepVGG_B3g4(pretrained=False, use_ssld=False, **kwargs): model = RepVGG( num_blocks=[4, 6, 16, 1], diff --git a/ppcls/arch/backbone/model_zoo/res2net.py b/ppcls/arch/backbone/model_zoo/res2net.py index 191cc849cb56012bdb0de88cb92558f64f2df2b9..87187075578ea312c13cd5e56cfa24fe8da0d8ce 100644 --- a/ppcls/arch/backbone/model_zoo/res2net.py +++ b/ppcls/arch/backbone/model_zoo/res2net.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1904.01169 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/res2net_vd.py b/ppcls/arch/backbone/model_zoo/res2net_vd.py index a375679803fa037672a0f2f741a9e0a517b996bc..511fbaa59e6ff5b4e5419edc084631f6e43873fa 100644 --- a/ppcls/arch/backbone/model_zoo/res2net_vd.py +++ b/ppcls/arch/backbone/model_zoo/res2net_vd.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1904.01169 & https://arxiv.org/abs/1812.01187 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/resnest.py b/ppcls/arch/backbone/model_zoo/resnest.py index 88eee8aacc9b678cea68cce1c3f54efa95578e8c..b4a978cc9e3fe68180348120be73deff229b77d2 100644 --- a/ppcls/arch/backbone/model_zoo/resnest.py +++ b/ppcls/arch/backbone/model_zoo/resnest.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/zhanghang1989/ResNeSt +# reference: https://arxiv.org/abs/2004.08955 from __future__ import absolute_import from __future__ import division diff --git a/ppcls/arch/backbone/model_zoo/resnet_vc.py b/ppcls/arch/backbone/model_zoo/resnet_vc.py index 6b972dc7bd18d9f2b2822e2915d06c6a88328822..1cbd0e9cec423ac71c35bd866f7037326d3350b6 100644 --- a/ppcls/arch/backbone/model_zoo/resnet_vc.py +++ b/ppcls/arch/backbone/model_zoo/resnet_vc.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1812.01187 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/resnext.py b/ppcls/arch/backbone/model_zoo/resnext.py index 1aef81144044fb055f66bd734e7c7c26aaef98d5..3ab21399fc1bd695accfd1c608536c9a2631f0b3 100644 --- a/ppcls/arch/backbone/model_zoo/resnext.py +++ b/ppcls/arch/backbone/model_zoo/resnext.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1611.05431 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/resnext101_wsl.py b/ppcls/arch/backbone/model_zoo/resnext101_wsl.py index e85e13388ab13586f8dac6e8bd42fe68bfc1ca52..bcf2bf542cf136fa61eda5a4d682ea63adc7f803 100644 --- a/ppcls/arch/backbone/model_zoo/resnext101_wsl.py +++ b/ppcls/arch/backbone/model_zoo/resnext101_wsl.py @@ -1,3 +1,19 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# reference: https://arxiv.org/abs/1805.00932 + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/resnext_vd.py b/ppcls/arch/backbone/model_zoo/resnext_vd.py index b2bd484f36ef62b41f8ed8daa0bf441ddd2bd815..d57e71e346514f779c0376541a3570665067a5f3 100644 --- a/ppcls/arch/backbone/model_zoo/resnext_vd.py +++ b/ppcls/arch/backbone/model_zoo/resnext_vd.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1611.05431 & https://arxiv.org/abs/1812.01187 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/rexnet.py b/ppcls/arch/backbone/model_zoo/rexnet.py index 1556a01c4c4dcb8b46931984e9ae744c68e8d9c9..d2f32af395698399d8314cf029429cb9eb641e0a 100644 --- a/ppcls/arch/backbone/model_zoo/rexnet.py +++ b/ppcls/arch/backbone/model_zoo/rexnet.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/2007.00992 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/se_resnet_vd.py b/ppcls/arch/backbone/model_zoo/se_resnet_vd.py index 205feec54a549e15d43b620a1d6666cb65fee2df..bded11fb9437e6a8d7d43ffc9dd923ca6a244772 100644 --- a/ppcls/arch/backbone/model_zoo/se_resnet_vd.py +++ b/ppcls/arch/backbone/model_zoo/se_resnet_vd.py @@ -11,6 +11,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1812.01187 & https://arxiv.org/abs/1709.01507 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/se_resnext.py b/ppcls/arch/backbone/model_zoo/se_resnext.py index 8b7149e26e020080c6c7d7a8e253da53397998e9..d1b985b42b20afeb4427a50d8124ce805e36e434 100644 --- a/ppcls/arch/backbone/model_zoo/se_resnext.py +++ b/ppcls/arch/backbone/model_zoo/se_resnext.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1611.05431 & https://arxiv.org/abs/1709.01507 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/se_resnext_vd.py b/ppcls/arch/backbone/model_zoo/se_resnext_vd.py index ef630259d3828c54f6089184326806ce7e7a5d13..78582f257bc0c890f942f8c5e32374e0de69d934 100644 --- a/ppcls/arch/backbone/model_zoo/se_resnext_vd.py +++ b/ppcls/arch/backbone/model_zoo/se_resnext_vd.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1611.05431 & https://arxiv.org/abs/1812.01187 & https://arxiv.org/abs/1709.01507 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/shufflenet_v2.py b/ppcls/arch/backbone/model_zoo/shufflenet_v2.py index d8bb69ffec96f622907146047c402d5a7d371171..b10249b7e2ea59bfa846c4fa3e09c5fbfe77b9ef 100644 --- a/ppcls/arch/backbone/model_zoo/shufflenet_v2.py +++ b/ppcls/arch/backbone/model_zoo/shufflenet_v2.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1807.11164 + from __future__ import absolute_import from __future__ import division from __future__ import print_function diff --git a/ppcls/arch/backbone/model_zoo/squeezenet.py b/ppcls/arch/backbone/model_zoo/squeezenet.py index 647cd2ea7c1bd71594a16774516ae3ce802a7d8b..3e8ea5d2a7f111c1cecd1374203d85817cecbe49 100644 --- a/ppcls/arch/backbone/model_zoo/squeezenet.py +++ b/ppcls/arch/backbone/model_zoo/squeezenet.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1709.01507 + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/swin_transformer.py b/ppcls/arch/backbone/model_zoo/swin_transformer.py index c783ec6806edf88625481ef800ac322ba8d71e57..877b7365998bce81489a89ab57a240deb66d45cc 100644 --- a/ppcls/arch/backbone/model_zoo/swin_transformer.py +++ b/ppcls/arch/backbone/model_zoo/swin_transformer.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/microsoft/Swin-Transformer +# reference: https://arxiv.org/abs/2103.14030 import numpy as np import paddle diff --git a/ppcls/arch/backbone/model_zoo/tnt.py b/ppcls/arch/backbone/model_zoo/tnt.py index 3ccb7575effb623457e92c9a79c2f891c10fbc71..c463ef2002101ddb3017164a64cbec9782ff2f4c 100644 --- a/ppcls/arch/backbone/model_zoo/tnt.py +++ b/ppcls/arch/backbone/model_zoo/tnt.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/huawei-noah/CV-Backbones/tree/master/tnt_pytorch +# reference: https://arxiv.org/abs/2103.00112 import math import numpy as np @@ -372,7 +373,7 @@ def _load_pretrained(pretrained, model, model_url, use_ssld=False): ) -def TNT_small(pretrained=False, **kwargs): +def TNT_small(pretrained=False, use_ssld=False, **kwargs): model = TNT(patch_size=16, embed_dim=384, in_dim=24, @@ -381,5 +382,6 @@ def TNT_small(pretrained=False, **kwargs): in_num_head=4, qkv_bias=False, **kwargs) - _load_pretrained(pretrained, model, MODEL_URLS["TNT_small"]) + _load_pretrained( + pretrained, model, MODEL_URLS["TNT_small"], use_ssld=use_ssld) return model diff --git a/ppcls/arch/backbone/model_zoo/van.py b/ppcls/arch/backbone/model_zoo/van.py new file mode 100644 index 0000000000000000000000000000000000000000..17ae5d69b8862173a84f60d684e2f8cac8c4791a --- /dev/null +++ b/ppcls/arch/backbone/model_zoo/van.py @@ -0,0 +1,318 @@ +# copyright (c) 2022 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Code was heavily based on https://github.com/Visual-Attention-Network/VAN-Classification +# reference: https://arxiv.org/abs/2202.09741 + +from functools import partial +import math +import paddle +import paddle.nn as nn +from paddle.nn.initializer import TruncatedNormal, Constant + +from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url + +MODEL_URLS = { + "VAN_tiny": "", # TODO +} + +__all__ = list(MODEL_URLS.keys()) + +trunc_normal_ = TruncatedNormal(std=.02) +zeros_ = Constant(value=0.) +ones_ = Constant(value=1.) + + +def drop_path(x, drop_prob=0., training=False): + """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). + the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper... + See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... + """ + if drop_prob == 0. or not training: + return x + keep_prob = paddle.to_tensor(1 - drop_prob) + shape = (paddle.shape(x)[0], ) + (1, ) * (x.ndim - 1) + random_tensor = keep_prob + paddle.rand(shape, dtype=x.dtype) + random_tensor = paddle.floor(random_tensor) # binarize + output = x.divide(keep_prob) * random_tensor + return output + + +class DropPath(nn.Layer): + """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). + """ + + def __init__(self, drop_prob=None): + super(DropPath, self).__init__() + self.drop_prob = drop_prob + + def forward(self, x): + return drop_path(x, self.drop_prob, self.training) + + +@paddle.jit.not_to_static +def swapdim(x, dim1, dim2): + a = list(range(len(x.shape))) + a[dim1], a[dim2] = a[dim2], a[dim1] + return x.transpose(a) + + +class Mlp(nn.Layer): + def __init__(self, + in_features, + hidden_features=None, + out_features=None, + act_layer=nn.GELU, + drop=0.): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Conv2D(in_features, hidden_features, 1) + self.dwconv = DWConv(hidden_features) + self.act = act_layer() + self.fc2 = nn.Conv2D(hidden_features, out_features, 1) + self.drop = nn.Dropout(drop) + + def forward(self, x): + x = self.fc1(x) + x = self.dwconv(x) + x = self.act(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +class LKA(nn.Layer): + def __init__(self, dim): + super().__init__() + self.conv0 = nn.Conv2D(dim, dim, 5, padding=2, groups=dim) + self.conv_spatial = nn.Conv2D( + dim, dim, 7, stride=1, padding=9, groups=dim, dilation=3) + self.conv1 = nn.Conv2D(dim, dim, 1) + + def forward(self, x): + attn = self.conv0(x) + attn = self.conv_spatial(attn) + attn = self.conv1(attn) + return x * attn + + +class Attention(nn.Layer): + def __init__(self, d_model): + super().__init__() + self.proj_1 = nn.Conv2D(d_model, d_model, 1) + self.activation = nn.GELU() + self.spatial_gating_unit = LKA(d_model) + self.proj_2 = nn.Conv2D(d_model, d_model, 1) + + def forward(self, x): + shorcut = x + x = self.proj_1(x) + x = self.activation(x) + x = self.spatial_gating_unit(x) + x = self.proj_2(x) + x = x + shorcut + return x + + +class Block(nn.Layer): + def __init__(self, + dim, + mlp_ratio=4., + drop=0., + drop_path=0., + act_layer=nn.GELU): + super().__init__() + self.norm1 = nn.BatchNorm2D(dim) + self.attn = Attention(dim) + self.drop_path = DropPath( + drop_path) if drop_path > 0. else nn.Identity() + self.norm2 = nn.BatchNorm2D(dim) + mlp_hidden_dim = int(dim * mlp_ratio) + self.mlp = Mlp(in_features=dim, + hidden_features=mlp_hidden_dim, + act_layer=act_layer, + drop=drop) + layer_scale_init_value = 1e-2 + self.layer_scale_1 = self.create_parameter( + shape=[dim, 1, 1], + default_initializer=Constant(value=layer_scale_init_value)) + self.layer_scale_2 = self.create_parameter( + shape=[dim, 1, 1], + default_initializer=Constant(value=layer_scale_init_value)) + + def forward(self, x): + x = x + self.drop_path(self.layer_scale_1 * self.attn(self.norm1(x))) + x = x + self.drop_path(self.layer_scale_2 * self.mlp(self.norm2(x))) + return x + + +class OverlapPatchEmbed(nn.Layer): + """ Image to Patch Embedding + """ + + def __init__(self, + img_size=224, + patch_size=7, + stride=4, + in_chans=3, + embed_dim=768): + super().__init__() + self.proj = nn.Conv2D( + in_chans, + embed_dim, + kernel_size=patch_size, + stride=stride, + padding=patch_size // 2) + self.norm = nn.BatchNorm2D(embed_dim) + + def forward(self, x): + x = self.proj(x) + _, _, H, W = x.shape + x = self.norm(x) + return x, H, W + + +class VAN(nn.Layer): + r""" VAN + A PaddlePaddle impl of : `Visual Attention Network` - + https://arxiv.org/pdf/2202.09741.pdf + """ + + def __init__(self, + img_size=224, + in_chans=3, + class_num=1000, + embed_dims=[64, 128, 256, 512], + mlp_ratios=[4, 4, 4, 4], + drop_rate=0., + drop_path_rate=0., + norm_layer=nn.LayerNorm, + depths=[3, 4, 6, 3], + num_stages=4, + flag=False): + super().__init__() + if flag == False: + self.class_num = class_num + self.depths = depths + self.num_stages = num_stages + + dpr = [x for x in paddle.linspace(0, drop_path_rate, sum(depths)) + ] # stochastic depth decay rule + cur = 0 + + for i in range(num_stages): + patch_embed = OverlapPatchEmbed( + img_size=img_size if i == 0 else img_size // (2**(i + 1)), + patch_size=7 if i == 0 else 3, + stride=4 if i == 0 else 2, + in_chans=in_chans if i == 0 else embed_dims[i - 1], + embed_dim=embed_dims[i]) + + block = nn.LayerList([ + Block( + dim=embed_dims[i], + mlp_ratio=mlp_ratios[i], + drop=drop_rate, + drop_path=dpr[cur + j]) for j in range(depths[i]) + ]) + norm = norm_layer(embed_dims[i]) + cur += depths[i] + + setattr(self, f"patch_embed{i + 1}", patch_embed) + setattr(self, f"block{i + 1}", block) + setattr(self, f"norm{i + 1}", norm) + + # classification head + self.head = nn.Linear(embed_dims[3], + class_num) if class_num > 0 else nn.Identity() + + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight) + if isinstance(m, nn.Linear) and m.bias is not None: + zeros_(m.bias) + elif isinstance(m, nn.LayerNorm): + zeros_(m.bias) + ones_(m.weight) + elif isinstance(m, nn.Conv2D): + fan_out = m._kernel_size[0] * m._kernel_size[1] * m._out_channels + fan_out //= m._groups + m.weight.set_value( + paddle.normal( + std=math.sqrt(2.0 / fan_out), shape=m.weight.shape)) + if m.bias is not None: + zeros_(m.bias) + + def forward_features(self, x): + B = x.shape[0] + + for i in range(self.num_stages): + patch_embed = getattr(self, f"patch_embed{i + 1}") + block = getattr(self, f"block{i + 1}") + norm = getattr(self, f"norm{i + 1}") + x, H, W = patch_embed(x) + for blk in block: + x = blk(x) + x = x.flatten(2) + x = swapdim(x, 1, 2) + x = norm(x) + if i != self.num_stages - 1: + x = x.reshape([B, H, W, x.shape[2]]).transpose([0, 3, 1, 2]) + + return x.mean(axis=1) + + def forward(self, x): + x = self.forward_features(x) + x = self.head(x) + + return x + + +class DWConv(nn.Layer): + def __init__(self, dim=768): + super().__init__() + self.dwconv = nn.Conv2D(dim, dim, 3, 1, 1, bias_attr=True, groups=dim) + + def forward(self, x): + x = self.dwconv(x) + return x + + +def _load_pretrained(pretrained, model, model_url, use_ssld=False): + if pretrained is False: + pass + elif pretrained is True: + load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld) + elif isinstance(pretrained, str): + load_dygraph_pretrain(model, pretrained) + else: + raise RuntimeError( + "pretrained type is not available. Please use `string` or `boolean` type." + ) + + +def VAN_tiny(pretrained=False, use_ssld=False, **kwargs): + model = VAN(embed_dims=[32, 64, 160, 256], + mlp_ratios=[8, 8, 4, 4], + norm_layer=partial( + nn.LayerNorm, epsilon=1e-6), + depths=[3, 3, 5, 2], + **kwargs) + _load_pretrained( + pretrained, model, MODEL_URLS["VAN_tiny"], use_ssld=use_ssld) + return model diff --git a/ppcls/arch/backbone/model_zoo/vision_transformer.py b/ppcls/arch/backbone/model_zoo/vision_transformer.py index c71c0262f7f967bba26eb48cd857ecbc827ea9b0..d3f149d232d644825d4ed2f8b51a47ad9f80335f 100644 --- a/ppcls/arch/backbone/model_zoo/vision_transformer.py +++ b/ppcls/arch/backbone/model_zoo/vision_transformer.py @@ -13,6 +13,7 @@ # limitations under the License. # Code was based on https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py +# reference: https://arxiv.org/abs/2010.11929 from collections.abc import Callable diff --git a/ppcls/arch/backbone/model_zoo/xception.py b/ppcls/arch/backbone/model_zoo/xception.py index 2b843788bf2f7a7b77501d3b0341573d25fb5ffb..966092b98779e1ec60bacd92b060e27549e2724d 100644 --- a/ppcls/arch/backbone/model_zoo/xception.py +++ b/ppcls/arch/backbone/model_zoo/xception.py @@ -1,3 +1,19 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# reference: https://arxiv.org/abs/1610.02357 + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/backbone/model_zoo/xception_deeplab.py b/ppcls/arch/backbone/model_zoo/xception_deeplab.py index c52769b37b501a77e5ce9146a0e991153b81d240..2a80d173684ce207211b0173f7bd6c58a7e3e30d 100644 --- a/ppcls/arch/backbone/model_zoo/xception_deeplab.py +++ b/ppcls/arch/backbone/model_zoo/xception_deeplab.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1706.05587 + import paddle from paddle import ParamAttr import paddle.nn as nn diff --git a/ppcls/arch/distill/afd_attention.py b/ppcls/arch/distill/afd_attention.py new file mode 100644 index 0000000000000000000000000000000000000000..63b094f313c08f0963a6afbe72361430cbaf789e --- /dev/null +++ b/ppcls/arch/distill/afd_attention.py @@ -0,0 +1,123 @@ +#copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +#Unless required by applicable law or agreed to in writing, software +#distributed under the License is distributed on an "AS IS" BASIS, +#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +#See the License for the specific language governing permissions and +#limitations under the License. + +import paddle.nn as nn +import paddle.nn.functional as F +import paddle +import numpy as np + + +class LinearBNReLU(nn.Layer): + def __init__(self, nin, nout): + super().__init__() + self.linear = nn.Linear(nin, nout) + self.bn = nn.BatchNorm1D(nout) + self.relu = nn.ReLU() + + def forward(self, x, relu=True): + if relu: + return self.relu(self.bn(self.linear(x))) + return self.bn(self.linear(x)) + + +def unique_shape(s_shapes): + n_s = [] + unique_shapes = [] + n = -1 + for s_shape in s_shapes: + if s_shape not in unique_shapes: + unique_shapes.append(s_shape) + n += 1 + n_s.append(n) + return n_s, unique_shapes + + +class LinearTransformTeacher(nn.Layer): + def __init__(self, qk_dim, t_shapes, keys): + super().__init__() + self.teacher_keys = keys + self.t_shapes = [[1] + t_i for t_i in t_shapes] + self.query_layer = nn.LayerList( + [LinearBNReLU(t_shape[1], qk_dim) for t_shape in self.t_shapes]) + + def forward(self, t_features_dict): + g_t = [t_features_dict[key] for key in self.teacher_keys] + bs = g_t[0].shape[0] + channel_mean = [f_t.mean(3).mean(2) for f_t in g_t] + spatial_mean = [] + for i in range(len(g_t)): + c, h, w = g_t[i].shape[1:] + spatial_mean.append(g_t[i].pow(2).mean(1).reshape([bs, h * w])) + query = paddle.stack( + [ + query_layer( + f_t, relu=False) + for f_t, query_layer in zip(channel_mean, self.query_layer) + ], + axis=1) + value = [F.normalize(f_s, axis=1) for f_s in spatial_mean] + return {"query": query, "value": value} + + +class LinearTransformStudent(nn.Layer): + def __init__(self, qk_dim, t_shapes, s_shapes, keys): + super().__init__() + self.student_keys = keys + self.t_shapes = [[1] + t_i for t_i in t_shapes] + self.s_shapes = [[1] + s_i for s_i in s_shapes] + self.t = len(self.t_shapes) + self.s = len(self.s_shapes) + self.qk_dim = qk_dim + self.n_t, self.unique_t_shapes = unique_shape(self.t_shapes) + self.relu = nn.ReLU() + self.samplers = nn.LayerList( + [Sample(t_shape) for t_shape in self.unique_t_shapes]) + self.key_layer = nn.LayerList([ + LinearBNReLU(s_shape[1], self.qk_dim) for s_shape in self.s_shapes + ]) + self.bilinear = LinearBNReLU(qk_dim, qk_dim * len(self.t_shapes)) + + def forward(self, s_features_dict): + g_s = [s_features_dict[key] for key in self.student_keys] + bs = g_s[0].shape[0] + channel_mean = [f_s.mean(3).mean(2) for f_s in g_s] + spatial_mean = [sampler(g_s, bs) for sampler in self.samplers] + + key = paddle.stack( + [ + key_layer(f_s) + for key_layer, f_s in zip(self.key_layer, channel_mean) + ], + axis=1).reshape([-1, self.qk_dim]) # Bs x h + bilinear_key = self.bilinear( + key, relu=False).reshape([bs, self.s, self.t, self.qk_dim]) + value = [F.normalize(s_m, axis=2) for s_m in spatial_mean] + return {"bilinear_key": bilinear_key, "value": value} + + +class Sample(nn.Layer): + def __init__(self, t_shape): + super().__init__() + self.t_N, self.t_C, self.t_H, self.t_W = t_shape + self.sample = nn.AdaptiveAvgPool2D((self.t_H, self.t_W)) + + def forward(self, g_s, bs): + g_s = paddle.stack( + [ + self.sample(f_s.pow(2).mean( + 1, keepdim=True)).reshape([bs, self.t_H * self.t_W]) + for f_s in g_s + ], + axis=1) + return g_s diff --git a/ppcls/arch/gears/__init__.py b/ppcls/arch/gears/__init__.py index 75ca41d8a1ed3d4276e53d67f56ba7238fbdccff..8757aa4aeb4a510857ca4dc1c60696b1d6e86a0b 100644 --- a/ppcls/arch/gears/__init__.py +++ b/ppcls/arch/gears/__init__.py @@ -17,13 +17,16 @@ from .cosmargin import CosMargin from .circlemargin import CircleMargin from .fc import FC from .vehicle_neck import VehicleNeck +from paddle.nn import Tanh +from .bnneck import BNNeck __all__ = ['build_gear'] def build_gear(config): support_dict = [ - 'ArcMargin', 'CosMargin', 'CircleMargin', 'FC', 'VehicleNeck' + 'ArcMargin', 'CosMargin', 'CircleMargin', 'FC', 'VehicleNeck', 'Tanh', + 'BNNeck' ] module_name = config.pop('name') assert module_name in support_dict, Exception( diff --git a/ppcls/arch/gears/arcmargin.py b/ppcls/arch/gears/arcmargin.py index 22cc76e1d954913d71db208c46037666e2b59086..6c72a71a248baf81bbbcc1bfea54965a9fb70eb5 100644 --- a/ppcls/arch/gears/arcmargin.py +++ b/ppcls/arch/gears/arcmargin.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1801.07698 + import paddle import paddle.nn as nn import math diff --git a/ppcls/arch/gears/bnneck.py b/ppcls/arch/gears/bnneck.py new file mode 100644 index 0000000000000000000000000000000000000000..c2f10c79f9c3862102f7b425c18018d2c4cce15e --- /dev/null +++ b/ppcls/arch/gears/bnneck.py @@ -0,0 +1,51 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import absolute_import, division, print_function + +import paddle +import paddle.nn as nn + +from ppcls.arch.utils import get_param_attr_dict + + +class BNNeck(nn.Layer): + def __init__(self, num_features, **kwargs): + super().__init__() + weight_attr = paddle.ParamAttr( + initializer=paddle.nn.initializer.Constant(value=1.0)) + bias_attr = paddle.ParamAttr( + initializer=paddle.nn.initializer.Constant(value=0.0), + trainable=False) + + if 'weight_attr' in kwargs: + weight_attr = get_param_attr_dict(kwargs['weight_attr']) + + bias_attr = None + if 'bias_attr' in kwargs: + bias_attr = get_param_attr_dict(kwargs['bias_attr']) + + self.feat_bn = nn.BatchNorm1D( + num_features, + momentum=0.9, + epsilon=1e-05, + weight_attr=weight_attr, + bias_attr=bias_attr) + + self.flatten = nn.Flatten() + + def forward(self, x): + x = self.flatten(x) + x = self.feat_bn(x) + return x diff --git a/ppcls/arch/gears/circlemargin.py b/ppcls/arch/gears/circlemargin.py index d1bce83cb36f007a3be83e5fc0e34ac52e9fe642..c04d6618bdbf570341400e190dc8589cde8ea028 100644 --- a/ppcls/arch/gears/circlemargin.py +++ b/ppcls/arch/gears/circlemargin.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/2002.10857 + import math import paddle import paddle.nn as nn diff --git a/ppcls/arch/gears/cosmargin.py b/ppcls/arch/gears/cosmargin.py index 578b64c2bc279bb0c3379ebc291f1711adc3ad86..d420c0ace0677457061d7e23e6a933e1bc12aa5f 100644 --- a/ppcls/arch/gears/cosmargin.py +++ b/ppcls/arch/gears/cosmargin.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +# reference: https://arxiv.org/abs/1801.09414 + import paddle import math import paddle.nn as nn diff --git a/ppcls/arch/gears/fc.py b/ppcls/arch/gears/fc.py index b32474195e1ada4cd0a17b493f68f65a242d82cd..279c5496e4aeeef86f1ebdafbdbfe7468391fa2d 100644 --- a/ppcls/arch/gears/fc.py +++ b/ppcls/arch/gears/fc.py @@ -19,16 +19,29 @@ from __future__ import print_function import paddle import paddle.nn as nn +from ppcls.arch.utils import get_param_attr_dict + class FC(nn.Layer): - def __init__(self, embedding_size, class_num): + def __init__(self, embedding_size, class_num, **kwargs): super(FC, self).__init__() self.embedding_size = embedding_size self.class_num = class_num + weight_attr = paddle.ParamAttr( initializer=paddle.nn.initializer.XavierNormal()) - self.fc = paddle.nn.Linear( - self.embedding_size, self.class_num, weight_attr=weight_attr) + if 'weight_attr' in kwargs: + weight_attr = get_param_attr_dict(kwargs['weight_attr']) + + bias_attr = None + if 'bias_attr' in kwargs: + bias_attr = get_param_attr_dict(kwargs['bias_attr']) + + self.fc = nn.Linear( + self.embedding_size, + self.class_num, + weight_attr=weight_attr, + bias_attr=bias_attr) def forward(self, input, label=None): out = self.fc(input) diff --git a/ppcls/arch/utils.py b/ppcls/arch/utils.py index 308475d7dbe7e4b9702a9e9e2eb3a0210da26e7a..785b7fbbe7e609e5314b549355165d83715bd48a 100644 --- a/ppcls/arch/utils.py +++ b/ppcls/arch/utils.py @@ -14,9 +14,11 @@ import six import types +import paddle from difflib import SequenceMatcher from . import backbone +from typing import Any, Dict, Union def get_architectures(): @@ -51,3 +53,47 @@ def similar_architectures(name='', names=[], thresh=0.1, topk=10): scores.sort(key=lambda x: x[1], reverse=True) similar_names = [names[s[0]] for s in scores[:min(topk, len(scores))]] return similar_names + + +def get_param_attr_dict(ParamAttr_config: Union[None, bool, Dict[str, Dict]] + ) -> Union[None, bool, paddle.ParamAttr]: + """parse ParamAttr from an dict + + Args: + ParamAttr_config (Union[None, bool, Dict[str, Dict]]): ParamAttr configure + + Returns: + Union[None, bool, paddle.ParamAttr]: Generated ParamAttr + """ + if ParamAttr_config is None: + return None + if isinstance(ParamAttr_config, bool): + return ParamAttr_config + ParamAttr_dict = {} + if 'initializer' in ParamAttr_config: + initializer_cfg = ParamAttr_config.get('initializer') + if 'name' in initializer_cfg: + initializer_name = initializer_cfg.pop('name') + ParamAttr_dict['initializer'] = getattr( + paddle.nn.initializer, initializer_name)(**initializer_cfg) + else: + raise ValueError(f"'name' must specified in initializer_cfg") + if 'learning_rate' in ParamAttr_config: + # NOTE: only support an single value now + learning_rate_value = ParamAttr_config.get('learning_rate') + if isinstance(learning_rate_value, (int, float)): + ParamAttr_dict['learning_rate'] = learning_rate_value + else: + raise ValueError( + f"learning_rate_value must be float or int, but got {type(learning_rate_value)}" + ) + if 'regularizer' in ParamAttr_config: + regularizer_cfg = ParamAttr_config.get('regularizer') + if 'name' in regularizer_cfg: + # L1Decay or L2Decay + regularizer_name = regularizer_cfg.pop('name') + ParamAttr_dict['regularizer'] = getattr( + paddle.regularizer, regularizer_name)(**regularizer_cfg) + else: + raise ValueError(f"'name' must specified in regularizer_cfg") + return paddle.ParamAttr(**ParamAttr_dict) diff --git a/ppcls/configs/DeepHash/DCH.yaml b/ppcls/configs/DeepHash/DCH.yaml new file mode 100644 index 0000000000000000000000000000000000000000..685ba447dada1ef46c58de044af2f1e0efcf99fc --- /dev/null +++ b/ppcls/configs/DeepHash/DCH.yaml @@ -0,0 +1,140 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output + device: gpu + save_interval: 15 + eval_during_train: True + eval_interval: 15 + epochs: 150 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + + #feature postprocess + feature_normalize: False + feature_binarize: "sign" + +# model architecture +Arch: + name: RecModel + infer_output_key: features + infer_add_softmax: False + + Backbone: + name: AlexNet + pretrained: True + class_num: 48 + +# loss function config for train/eval process +Loss: + Train: + - DCHLoss: + weight: 1.0 + gamma: 20.0 + _lambda: 0.1 + n_class: 10 + Eval: + - DCHLoss: + weight: 1.0 + gamma: 20.0 + _lambda: 0.1 + n_class: 10 + +Optimizer: + name: SGD + lr: + name: Piecewise + learning_rate: 0.005 + decay_epochs: [200] + values: [0.005, 0.0005] + regularizer: + name: 'L2' + coeff: 0.00001 + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - mAP: {} + - Recallk: + topk: [1, 5] diff --git a/ppcls/configs/DeepHash/DSHSD.yaml b/ppcls/configs/DeepHash/DSHSD.yaml new file mode 100644 index 0000000000000000000000000000000000000000..77cc7dc1d0147fb2275f5b2a27d867c39aacd2bc --- /dev/null +++ b/ppcls/configs/DeepHash/DSHSD.yaml @@ -0,0 +1,141 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output + device: gpu + save_interval: 15 + eval_during_train: True + eval_interval: 15 + epochs: 150 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + + #feature postprocess + feature_normalize: False + feature_binarize: "sign" + +# model architecture +Arch: + name: RecModel + infer_output_key: features + infer_add_softmax: False + + Backbone: + name: AlexNet + pretrained: True + class_num: 48 + Neck: + name: Tanh + Head: + name: FC + class_num: 10 + embedding_size: 48 + +# loss function config for train/eval process +Loss: + Train: + - DSHSDLoss: + weight: 1.0 + alpha: 0.05 + Eval: + - DSHSDLoss: + weight: 1.0 + alpha: 0.05 + +Optimizer: + name: Adam + beta1: 0.9 + beta2: 0.999 + lr: + name: Piecewise + learning_rate: 0.00001 + decay_epochs: [200] + values: [0.00001, 0.000001] + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - mAP: {} + - Recallk: + topk: [1, 5] diff --git a/ppcls/configs/DeepHash/LCDSH.yaml b/ppcls/configs/DeepHash/LCDSH.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fe9243f91ab5a9f0cf68f41a57b8d7ab16d569d6 --- /dev/null +++ b/ppcls/configs/DeepHash/LCDSH.yaml @@ -0,0 +1,137 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output + device: gpu + save_interval: 15 + eval_during_train: True + eval_interval: 15 + epochs: 150 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + eval_mode: retrieval + use_dali: False + to_static: False + + #feature postprocess + feature_normalize: False + feature_binarize: "sign" + +# model architecture +Arch: + name: RecModel + infer_output_key: features + infer_add_softmax: False + + Backbone: + name: AlexNet + pretrained: True + class_num: 48 + +# loss function config for train/eval process +Loss: + Train: + - LCDSHLoss: + weight: 1.0 + _lambda: 3 + n_class: 10 + Eval: + - LCDSHLoss: + weight: 1.0 + _lambda: 3 + n_class: 10 + +Optimizer: + name: Adam + beta1: 0.9 + beta2: 0.999 + lr: + name: Piecewise + learning_rate: 0.00001 + decay_epochs: [200] + values: [0.00001, 0.000001] + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + Query: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/test_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: ImageNetDataset + image_root: ./dataset/CIFAR10/ + cls_label_path: ./dataset/CIFAR10/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - mAP: {} + - Recallk: + topk: [1, 5] diff --git a/ppcls/configs/GeneralRecognition/Gallery2FC_PPLCNet_x2_5.yaml b/ppcls/configs/GeneralRecognition/Gallery2FC_PPLCNet_x2_5.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fbaefdcf547f5f1a43d5e7db5079e8c5bc3704c8 --- /dev/null +++ b/ppcls/configs/GeneralRecognition/Gallery2FC_PPLCNet_x2_5.yaml @@ -0,0 +1,51 @@ +# global configs +Global: + pretrained_model: ./pretrained/general_PPLCNet_x2_5_pretrained_v1.0_quant + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference/general_PPLCNet_x2_5_quant/inference + +# for quantizaiton or prune model +Slim: + ## for prune + quant: + name: pact + +# model architecture +Arch: + name: RecModel + + Backbone: + name: PPLCNet_x2_5 + pretrained: False + use_ssld: True + BackboneStopLayer: + name: "flatten" + Neck: + name: FC + embedding_size: 1280 + class_num: 512 + Head: + name: ArcMargin + embedding_size: 512 + class_num: 185341 + margin: 0.2 + scale: 30 + +# indexing engine config +IndexProcess: + image_root: "./drink_dataset_v1.0/gallery/" + data_file: "./drink_dataset_v1.0/gallery/drink_label.txt" + delimiter: "\t" + batch_size: 2 + transform_ops: + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [ 0.485, 0.456, 0.406 ] + std: [ 0.229, 0.224, 0.225 ] + order: '' + - ToCHWImage: diff --git a/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_224.yaml b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_224.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a7697840ea8f1f29bed5d9e2c2226ea18f4e421d --- /dev/null +++ b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_224.yaml @@ -0,0 +1,162 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + + image_shape: [3, 224, 224] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + +# model architecture +Arch: + name: CSWinTransformer_base_224 + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.05 + no_weight_decay_name: pos_embed cls_token .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 2.5e-4 + eta_min: 2.5e-6 + warmup_epoch: 20 + warmup_start_lr: 2.5e-7 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + interpolation: bicubic + backend: pil + - RandFlipImage: + flip_code: 1 + - TimmAutoAugment: + config_str: rand-m9-mstd0.5-inc1 + interpolation: bicubic + img_size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - RandomErasing: + EPSILON: 0.25 + sl: 0.02 + sh: 1.0/3.0 + r1: 0.3 + attempt: 10 + use_log_aspect: True + mode: pixel + batch_transform_ops: + - OpSampler: + MixupOperator: + alpha: 0.8 + prob: 0.5 + CutmixOperator: + alpha: 1.0 + prob: 0.5 + + sampler: + name: DistributedBatchSampler + batch_size: 32 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Eval: + - TopkAcc: + topk: [1, 5] + diff --git a/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_384.yaml b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_384.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a7100289c06b94f211dd3fea6cd0b8f2548b8244 --- /dev/null +++ b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_384.yaml @@ -0,0 +1,161 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + + image_shape: [3, 384, 384] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + +# model architecture +Arch: + name: CSWinTransformer_base_384 + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.05 + no_weight_decay_name: pos_embed cls_token .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 1.25e-4 + eta_min: 1.25e-6 + warmup_epoch: 20 + warmup_start_lr: 1.25e-7 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 384 + interpolation: bicubic + backend: pil + - RandFlipImage: + flip_code: 1 + - TimmAutoAugment: + config_str: rand-m9-mstd0.5-inc1 + interpolation: bicubic + img_size: 384 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - RandomErasing: + EPSILON: 0.25 + sl: 0.02 + sh: 1.0/3.0 + r1: 0.3 + attempt: 10 + use_log_aspect: True + mode: pixel + batch_transform_ops: + - OpSampler: + MixupOperator: + alpha: 0.8 + prob: 0.5 + CutmixOperator: + alpha: 1.0 + prob: 0.5 + + sampler: + name: DistributedBatchSampler + batch_size: 16 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 384 + interpolation: bicubic + backend: pil + - CropImage: + size: 384 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 384 + interpolation: bicubic + backend: pil + - CropImage: + size: 384 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_224.yaml b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_224.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7c96343df5d00f8c7ceeba0f1dabec8ecccdbc57 --- /dev/null +++ b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_224.yaml @@ -0,0 +1,161 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + + image_shape: [3, 224, 224] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + +# model architecture +Arch: + name: CSWinTransformer_large_224 + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.05 + no_weight_decay_name: pos_embed cls_token .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 2.5e-4 + eta_min: 2.5e-6 + warmup_epoch: 20 + warmup_start_lr: 2.5e-7 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + interpolation: bicubic + backend: pil + - RandFlipImage: + flip_code: 1 + - TimmAutoAugment: + config_str: rand-m9-mstd0.5-inc1 + interpolation: bicubic + img_size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - RandomErasing: + EPSILON: 0.25 + sl: 0.02 + sh: 1.0/3.0 + r1: 0.3 + attempt: 10 + use_log_aspect: True + mode: pixel + batch_transform_ops: + - OpSampler: + MixupOperator: + alpha: 0.8 + prob: 0.5 + CutmixOperator: + alpha: 1.0 + prob: 0.5 + + sampler: + name: DistributedBatchSampler + batch_size: 32 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_384.yaml b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_384.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4b682fec60569cd16e6cf81d093d016761f37615 --- /dev/null +++ b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_384.yaml @@ -0,0 +1,161 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + + image_shape: [3, 384, 384] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + +# model architecture +Arch: + name: CSWinTransformer_large_384 + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.05 + no_weight_decay_name: pos_embed cls_token .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 6.25e-5 + eta_min: 6.25e-7 + warmup_epoch: 20 + warmup_start_lr: 6.25e-8 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 384 + interpolation: bicubic + backend: pil + - RandFlipImage: + flip_code: 1 + - TimmAutoAugment: + config_str: rand-m9-mstd0.5-inc1 + interpolation: bicubic + img_size: 384 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - RandomErasing: + EPSILON: 0.25 + sl: 0.02 + sh: 1.0/3.0 + r1: 0.3 + attempt: 10 + use_log_aspect: True + mode: pixel + batch_transform_ops: + - OpSampler: + MixupOperator: + alpha: 0.8 + prob: 0.5 + CutmixOperator: + alpha: 1.0 + prob: 0.5 + + sampler: + name: DistributedBatchSampler + batch_size: 8 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 384 + interpolation: bicubic + backend: pil + - CropImage: + size: 384 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 384 + interpolation: bicubic + backend: pil + - CropImage: + size: 384 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_small_224.yaml b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_small_224.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a191f4160fd2922974a14c379a8010f90a698b6c --- /dev/null +++ b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_small_224.yaml @@ -0,0 +1,161 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + + image_shape: [3, 224, 224] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + +# model architecture +Arch: + name: CSWinTransformer_small_224 + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.05 + no_weight_decay_name: pos_embed cls_token .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 5e-4 + eta_min: 5e-6 + warmup_epoch: 20 + warmup_start_lr: 5e-7 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + interpolation: bicubic + backend: pil + - RandFlipImage: + flip_code: 1 + - TimmAutoAugment: + config_str: rand-m9-mstd0.5-inc1 + interpolation: bicubic + img_size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - RandomErasing: + EPSILON: 0.25 + sl: 0.02 + sh: 1.0/3.0 + r1: 0.3 + attempt: 10 + use_log_aspect: True + mode: pixel + batch_transform_ops: + - OpSampler: + MixupOperator: + alpha: 0.8 + prob: 0.5 + CutmixOperator: + alpha: 1.0 + prob: 0.5 + + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_tiny_224.yaml b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_tiny_224.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3a2be2837891639604cbea15b7179adfb5cda0c7 --- /dev/null +++ b/ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_tiny_224.yaml @@ -0,0 +1,161 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + + image_shape: [3, 224, 224] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + +# model architecture +Arch: + name: CSWinTransformer_tiny_224 + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.05 + no_weight_decay_name: pos_embed cls_token .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 1e-3 + eta_min: 1e-5 + warmup_epoch: 20 + warmup_start_lr: 1e-6 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + interpolation: bicubic + backend: pil + - RandFlipImage: + flip_code: 1 + - TimmAutoAugment: + config_str: rand-m9-mstd0.5-inc1 + interpolation: bicubic + img_size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - RandomErasing: + EPSILON: 0.25 + sl: 0.02 + sh: 1.0/3.0 + r1: 0.3 + attempt: 10 + use_log_aspect: True + mode: pixel + batch_transform_ops: + - OpSampler: + MixupOperator: + alpha: 0.8 + prob: 0.5 + CutmixOperator: + alpha: 1.0 + prob: 0.5 + + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_224.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_224.yaml index 979a04a381c67debd81ffc6a509664de8cd71ccd..8c3cc4c3413729e2a5530de23105619ae76e0a15 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_224.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_224.yaml @@ -40,11 +40,12 @@ Optimizer: no_weight_decay_name: norm cls_token pos_embed dist_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_384.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_384.yaml index 859f57d72b0ee1261e3d06bb1c8dec6e601faada..0b8c2e808c94f09f9a6283a6baa605bd7ac45d85 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_384.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_base_distilled_patch16_384.yaml @@ -40,11 +40,12 @@ Optimizer: no_weight_decay_name: norm cls_token pos_embed dist_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml index 3cdd10202ee8275eeb9d1d1535e414007559fd6d..938916caa3937b2f1ef556e98c3adf9313b4c7b9 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml @@ -40,11 +40,12 @@ Optimizer: no_weight_decay_name: norm cls_token pos_embed dist_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml index 88a8fbae9394d81438f96a99cc9ac9da243cfba8..4cbe6ffded134bbe52656879b5105f3676c44b64 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml @@ -40,11 +40,12 @@ Optimizer: no_weight_decay_name: norm cls_token pos_embed dist_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_small_distilled_patch16_224.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_small_distilled_patch16_224.yaml index 54d962e689b95d78c56b7c0bd8a0fd44a379364f..d5ba0cee78de68ecccb35ffcbf099ddbdad3271d 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_small_distilled_patch16_224.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_small_distilled_patch16_224.yaml @@ -41,10 +41,10 @@ Optimizer: one_dim_param_no_weight_decay: True lr: name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml index 05c3ac1f36ad79ffa141eb86d4fceabbe28da98b..a167c896e8be7746d9896cfcd9d3d3a8e7671908 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml @@ -40,11 +40,12 @@ Optimizer: no_weight_decay_name: norm cls_token pos_embed dist_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_tiny_distilled_patch16_224.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_tiny_distilled_patch16_224.yaml index f6661761391d65c35152ee222eb7b6cf3273de47..319e17025d758eadce16001863312f773410104c 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_tiny_distilled_patch16_224.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_tiny_distilled_patch16_224.yaml @@ -40,11 +40,12 @@ Optimizer: no_weight_decay_name: norm cls_token pos_embed dist_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml b/ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml index 647050a77b181f0ac5ccbf473130415d6b957513..1234d79b6ba68186466edd7c2d1ea4f6bc61eba9 100644 --- a/ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml +++ b/ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml @@ -40,11 +40,12 @@ Optimizer: no_weight_decay_name: norm cls_token pos_embed dist_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 1e-3 - eta_min: 1e-5 + learning_rate: 2e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval DataLoader: diff --git a/ppcls/configs/ImageNet/Distillation/mv3_large_x1_0_distill_mv3_small_x1_0.yaml b/ppcls/configs/ImageNet/Distillation/mv3_large_x1_0_distill_mv3_small_x1_0.yaml index 9612be4e61bcdda39d568d0422a5b741a96b9d20..b230f11cbde78e195355d00a7b042b0d9e6a4026 100644 --- a/ppcls/configs/ImageNet/Distillation/mv3_large_x1_0_distill_mv3_small_x1_0.yaml +++ b/ppcls/configs/ImageNet/Distillation/mv3_large_x1_0_distill_mv3_small_x1_0.yaml @@ -18,7 +18,7 @@ Global: # model architecture Arch: name: "DistillationModel" - class_num: 1000 + class_num: &class_num 1000 # if not null, its lengths should be same as models pretrained_list: # if not null, its lengths should be same as models @@ -28,11 +28,13 @@ Arch: models: - Teacher: name: MobileNetV3_large_x1_0 + class_num: *class_num pretrained: True use_ssld: True dropout_prob: null - Student: name: MobileNetV3_small_x1_0 + class_num: *class_num pretrained: False dropout_prob: null @@ -47,9 +49,8 @@ Loss: model_name_pairs: - ["Student", "Teacher"] Eval: - - DistillationGTCELoss: + - CELoss: weight: 1.0 - model_names: ["Student"] Optimizer: diff --git a/ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_afd.yaml b/ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_afd.yaml new file mode 100644 index 0000000000000000000000000000000000000000..000cb9add132c0231d72d47e2947d4397c380d61 --- /dev/null +++ b/ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_afd.yaml @@ -0,0 +1,200 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: "./output/" + device: "gpu" + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 100 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: "./inference" + +# model architecture +Arch: + name: "DistillationModel" + # if not null, its lengths should be same as models + pretrained_list: + # if not null, its lengths should be same as models + freeze_params_list: + models: + - Teacher: + name: AttentionModel + pretrained_list: + freeze_params_list: + - True + - False + models: + - ResNet34: + name: ResNet34 + pretrained: True + return_patterns: &t_keys ["blocks[0]", "blocks[1]", "blocks[2]", "blocks[3]", + "blocks[4]", "blocks[5]", "blocks[6]", "blocks[7]", + "blocks[8]", "blocks[9]", "blocks[10]", "blocks[11]", + "blocks[12]", "blocks[13]", "blocks[14]", "blocks[15]"] + - LinearTransformTeacher: + name: LinearTransformTeacher + qk_dim: 128 + keys: *t_keys + t_shapes: &t_shapes [[64, 56, 56], [64, 56, 56], [64, 56, 56], [128, 28, 28], + [128, 28, 28], [128, 28, 28], [128, 28, 28], [256, 14, 14], + [256, 14, 14], [256, 14, 14], [256, 14, 14], [256, 14, 14], + [256, 14, 14], [512, 7, 7], [512, 7, 7], [512, 7, 7]] + + - Student: + name: AttentionModel + pretrained_list: + freeze_params_list: + - False + - False + models: + - ResNet18: + name: ResNet18 + pretrained: False + return_patterns: &s_keys ["blocks[0]", "blocks[1]", "blocks[2]", "blocks[3]", + "blocks[4]", "blocks[5]", "blocks[6]", "blocks[7]"] + - LinearTransformStudent: + name: LinearTransformStudent + qk_dim: 128 + keys: *s_keys + s_shapes: &s_shapes [[64, 56, 56], [64, 56, 56], [128, 28, 28], [128, 28, 28], + [256, 14, 14], [256, 14, 14], [512, 7, 7], [512, 7, 7]] + t_shapes: *t_shapes + + infer_model_name: "Student" + + +# loss function config for traing/eval process +Loss: + Train: + - DistillationGTCELoss: + weight: 1.0 + model_names: ["Student"] + key: logits + - DistillationKLDivLoss: + weight: 0.9 + model_name_pairs: [["Student", "Teacher"]] + temperature: 4 + key: logits + - AFDLoss: + weight: 50.0 + model_name_pair: ["Student", "Teacher"] + student_keys: ["bilinear_key", "value"] + teacher_keys: ["query", "value"] + s_shapes: *s_shapes + t_shapes: *t_shapes + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Momentum + momentum: 0.9 + weight_decay: 1e-4 + lr: + name: MultiStepDecay + learning_rate: 0.1 + milestones: [30, 60, 90] + step_each_epoch: 1 + gamma: 0.1 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: "./dataset/ILSVRC2012/" + cls_label_path: "./dataset/ILSVRC2012/train_list.txt" + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + interpolation: bicubic + backend: pil + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: True + loader: + num_workers: 8 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: "./dataset/ILSVRC2012/" + cls_label_path: "./dataset/ILSVRC2012/val_list.txt" + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 256 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: "docs/images/inference_deployment/whl_demo.jpg" + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 256 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: DistillationPostProcess + func: Topk + topk: 5 + class_id_map_file: "ppcls/utils/imagenet1k_label_list.txt" + +Metric: + Train: + - DistillationTopkAcc: + model_key: "Student" + topk: [1, 5] + Eval: + - DistillationTopkAcc: + model_key: "Student" + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml b/ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8efa5c054aed46380387c1df905742982a9c6c7d --- /dev/null +++ b/ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml @@ -0,0 +1,155 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: "./output/" + device: "gpu" + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 100 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: "./inference" + +# model architecture +Arch: + name: "DistillationModel" + # if not null, its lengths should be same as models + pretrained_list: + # if not null, its lengths should be same as models + freeze_params_list: + - True + - False + models: + - Teacher: + name: ResNet34 + pretrained: True + + - Student: + name: ResNet18 + pretrained: False + + infer_model_name: "Student" + + +# loss function config for traing/eval process +Loss: + Train: + - DistillationGTCELoss: + weight: 1.0 + model_names: ["Student"] + - DistillationDKDLoss: + weight: 1.0 + model_name_pairs: [["Student", "Teacher"]] + temperature: 1 + alpha: 1.0 + beta: 1.0 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: Momentum + momentum: 0.9 + weight_decay: 1e-4 + lr: + name: MultiStepDecay + learning_rate: 0.2 + milestones: [30, 60, 90] + step_each_epoch: 1 + gamma: 0.1 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: "./dataset/ILSVRC2012/" + cls_label_path: "./dataset/ILSVRC2012/train_list.txt" + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 8 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: "./dataset/ILSVRC2012/" + cls_label_path: "./dataset/ILSVRC2012/val_list.txt" + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 0.00392157 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: "docs/images/inference_deployment/whl_demo.jpg" + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - ToCHWImage: + PostProcess: + name: DistillationPostProcess + func: Topk + topk: 5 + class_id_map_file: "ppcls/utils/imagenet1k_label_list.txt" + +Metric: + Train: + - DistillationTopkAcc: + model_key: "Student" + topk: [1, 5] + Eval: + - DistillationTopkAcc: + model_key: "Student" + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/MobileViT/MobileViT_S.yaml b/ppcls/configs/ImageNet/MobileViT/MobileViT_S.yaml new file mode 100644 index 0000000000000000000000000000000000000000..becfef460701681d8dfbd3fabe9ec6d24c02e769 --- /dev/null +++ b/ppcls/configs/ImageNet/MobileViT/MobileViT_S.yaml @@ -0,0 +1,139 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 256, 256] + save_inference_dir: ./inference + use_dali: False +# model architecture +Arch: + name: MobileViT_S + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.01 + no_weight_decay_name: .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 0.002 + eta_min: 0.0002 + warmup_epoch: 5 + warmup_start_lr: 0.0002 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 256 + interpolation: bilinear + backend: pil + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 8 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: False + channel_first: False + - ResizeImage: + resize_short: 292 + interpolation: bilinear + backend: pil + - CropImage: + size: 256 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 292 + - CropImage: + size: 256 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Train: + - TopkAcc: + topk: [1, 5] + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/MobileViT/MobileViT_XS.yaml b/ppcls/configs/ImageNet/MobileViT/MobileViT_XS.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6cd152a134075455f88b00de1e15719d5e8903b0 --- /dev/null +++ b/ppcls/configs/ImageNet/MobileViT/MobileViT_XS.yaml @@ -0,0 +1,139 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 256, 256] + save_inference_dir: ./inference + use_dali: False +# model architecture +Arch: + name: MobileViT_XS + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.01 + no_weight_decay_name: .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 0.002 + eta_min: 0.0002 + warmup_epoch: 5 + warmup_start_lr: 0.0002 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 256 + interpolation: bilinear + backend: pil + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 8 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: False + channel_first: False + - ResizeImage: + resize_short: 292 + interpolation: bilinear + backend: pil + - CropImage: + size: 256 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 292 + - CropImage: + size: 256 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Train: + - TopkAcc: + topk: [1, 5] + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/MobileViT/MobileViT_XXS.yaml b/ppcls/configs/ImageNet/MobileViT/MobileViT_XXS.yaml new file mode 100644 index 0000000000000000000000000000000000000000..076b9f363a239588e1edc60cf744a70190fabedb --- /dev/null +++ b/ppcls/configs/ImageNet/MobileViT/MobileViT_XXS.yaml @@ -0,0 +1,139 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 256, 256] + save_inference_dir: ./inference + use_dali: False +# model architecture +Arch: + name: MobileViT_XXS + class_num: 1000 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.01 + no_weight_decay_name: .bias norm + one_dim_param_no_weight_decay: True + lr: + # for 8 cards + name: Cosine + learning_rate: 0.002 + eta_min: 0.0002 + warmup_epoch: 5 + warmup_start_lr: 0.0002 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 256 + interpolation: bilinear + backend: pil + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: True + loader: + num_workers: 8 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: False + channel_first: False + - ResizeImage: + resize_short: 292 + interpolation: bilinear + backend: pil + - CropImage: + size: 256 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 292 + - CropImage: + size: 256 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.0, 0.0, 0.0] + std: [1.0, 1.0, 1.0] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Train: + - TopkAcc: + topk: [1, 5] + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml index 6c0854cb4a3aed82f8e8897d26e5ad964ceb9cc9..27fc20b99961b29e9ddbcb58363b495f199b8aec 100644 --- a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml +++ b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B0.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: pos_embed1 pos_embed2 pos_embed3 pos_embed4 cls_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 5e-6 + learning_rate: 1e-3 + eta_min: 1e-5 warmup_epoch: 20 - warmup_start_lr: 5e-7 + warmup_start_lr: 1e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml index 42134c74006042a81a1bd4c9181105034fa56ae0..20fa39773f66d0ffff2786a031f86156c5fc5c41 100644 --- a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml +++ b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B1.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: pos_embed1 pos_embed2 pos_embed3 pos_embed4 cls_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 5e-6 + learning_rate: 1e-3 + eta_min: 1e-5 warmup_epoch: 20 - warmup_start_lr: 5e-7 + warmup_start_lr: 1e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml index 4d0d5a432dba117fcffc98b2c3f32afcbc7a0969..cda94496e34747468ac3dfe0b474478c0d30cae6 100644 --- a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml +++ b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: pos_embed1 pos_embed2 pos_embed3 pos_embed4 cls_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 5e-6 + learning_rate: 1e-3 + eta_min: 1e-5 warmup_epoch: 20 - warmup_start_lr: 5e-7 + warmup_start_lr: 1e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml index a5feb260bfdb570a8b0ae5791b74697051546e61..2d48178f05c19bcca34da95087129893fc574bd0 100644 --- a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml +++ b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: pos_embed1 pos_embed2 pos_embed3 pos_embed4 cls_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 5e-6 + learning_rate: 1e-3 + eta_min: 1e-5 warmup_epoch: 20 - warmup_start_lr: 5e-7 + warmup_start_lr: 1e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml index be300aca61832f9886daccb6786af770cc5fcb46..581a7060549607248f967e54b60956722bcb4be0 100644 --- a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml +++ b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B3.yaml @@ -44,11 +44,12 @@ Optimizer: no_weight_decay_name: pos_embed1 pos_embed2 pos_embed3 pos_embed4 cls_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 5e-6 + learning_rate: 1e-3 + eta_min: 1e-5 warmup_epoch: 20 - warmup_start_lr: 5e-7 + warmup_start_lr: 1e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml index b6a89533949e472ffea26e2b18ae1ea174288afa..92da84d1ef00430a1aab66fb33f84e8016ab2c0c 100644 --- a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml +++ b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B4.yaml @@ -44,11 +44,12 @@ Optimizer: no_weight_decay_name: pos_embed1 pos_embed2 pos_embed3 pos_embed4 cls_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 5e-6 + learning_rate: 1e-3 + eta_min: 1e-5 warmup_epoch: 20 - warmup_start_lr: 5e-7 + warmup_start_lr: 1e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml index 9d36b28078ff2472913da4cc3101ad39779d6a3d..4bb2449a40bdd605b7d36359437b39324e5b1772 100644 --- a/ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml +++ b/ppcls/configs/ImageNet/PVTV2/PVT_V2_B5.yaml @@ -44,11 +44,12 @@ Optimizer: no_weight_decay_name: pos_embed1 pos_embed2 pos_embed3 pos_embed4 cls_token one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 5e-6 + learning_rate: 1e-3 + eta_min: 1e-5 warmup_epoch: 20 - warmup_start_lr: 5e-7 + warmup_start_lr: 1e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml b/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml index 53e9ae21330e20d20813d0d7d8d4880c521e6529..4b87cb411a2f1c8d56329f1e2bbea7be82cc235b 100644 --- a/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml +++ b/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1.yaml @@ -10,9 +10,8 @@ Global: epochs: 120 print_batch_step: 10 use_visualdl: False - image_channel: &image_channel 4 # used for static mode and model export - image_shape: [*image_channel, 224, 224] + image_shape: [3, 224, 224] save_inference_dir: ./inference # training model under @to_static to_static: False @@ -29,8 +28,6 @@ AMP: Arch: name: ResNet50 class_num: 1000 - input_image_channel: *image_channel - data_format: "NHWC" # loss function config for traing/eval process Loss: @@ -76,7 +73,6 @@ DataLoader: mean: [0.485, 0.456, 0.406] std: [0.229, 0.224, 0.225] order: '' - channel_num: *image_channel sampler: name: DistributedBatchSampler @@ -105,7 +101,6 @@ DataLoader: mean: [0.485, 0.456, 0.406] std: [0.229, 0.224, 0.225] order: '' - channel_num: *image_channel sampler: name: DistributedBatchSampler batch_size: 64 @@ -131,7 +126,6 @@ Infer: mean: [0.485, 0.456, 0.406] std: [0.229, 0.224, 0.225] order: '' - channel_num: *image_channel - ToCHWImage: PostProcess: name: Topk diff --git a/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1_ultra.yaml b/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1_ultra.yaml new file mode 100644 index 0000000000000000000000000000000000000000..53e9ae21330e20d20813d0d7d8d4880c521e6529 --- /dev/null +++ b/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1_ultra.yaml @@ -0,0 +1,147 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 120 + print_batch_step: 10 + use_visualdl: False + image_channel: &image_channel 4 + # used for static mode and model export + image_shape: [*image_channel, 224, 224] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + use_dali: True + +# mixed precision training +AMP: + scale_loss: 128.0 + use_dynamic_loss_scaling: True + # O1: mixed fp16 + level: O1 + +# model architecture +Arch: + name: ResNet50 + class_num: 1000 + input_image_channel: *image_channel + data_format: "NHWC" + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + Eval: + - CELoss: + weight: 1.0 + + +Optimizer: + name: Momentum + momentum: 0.9 + multi_precision: True + lr: + name: Piecewise + learning_rate: 0.1 + decay_epochs: [30, 60, 90] + values: [0.1, 0.01, 0.001, 0.0001] + regularizer: + name: 'L2' + coeff: 0.0001 + + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + - RandFlipImage: + flip_code: 1 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + channel_num: *image_channel + + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + channel_num: *image_channel + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 256 + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + channel_num: *image_channel + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Train: + - TopkAcc: + topk: [1, 5] + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O2.yaml b/ppcls/configs/ImageNet/ResNet/ResNet50_amp_O2_ultra.yaml similarity index 100% rename from ppcls/configs/ImageNet/ResNet/ResNet50_amp_O2.yaml rename to ppcls/configs/ImageNet/ResNet/ResNet50_amp_O2_ultra.yaml diff --git a/ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d_amp_O2.yaml b/ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d_amp_O2_ultra.yaml similarity index 97% rename from ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d_amp_O2.yaml rename to ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d_amp_O2_ultra.yaml index da005d32911fdb942aa600b39937b660642f27f7..af987ed7f59ff9c9576d4fb417c48e112afa3986 100644 --- a/ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d_amp_O2.yaml +++ b/ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d_amp_O2_ultra.yaml @@ -34,10 +34,10 @@ Loss: # mixed precision training AMP: - scale_loss: 128.0 - use_dynamic_loss_scaling: True - # O2: pure fp16 - level: O2 + scale_loss: 128.0 + use_dynamic_loss_scaling: True + # O2: pure fp16 + level: O2 Optimizer: name: Momentum diff --git a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml index 4dd0ac4cfbe3840e39ca46b67b539c3c1a9cf146..afc3fdcd263de26e4864ecafbe46db7afd575ddf 100644 --- a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml +++ b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml @@ -41,11 +41,12 @@ Optimizer: no_weight_decay_name: absolute_pos_embed relative_position_bias_table .bias norm one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 20 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml index a42dea1f94b2c34b2d34497861cc8daf6ec757b1..4920fae6c4bab15f16d908d112a618da42aa9b35 100644 --- a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml +++ b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml @@ -41,11 +41,12 @@ Optimizer: no_weight_decay_name: absolute_pos_embed relative_position_bias_table .bias norm one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 20 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml index 36b5e5e38ecc2b70fc82b93dfd8a761c3e40d0b0..a6dd74267eaab84d919ff47d979d4ed863520ff8 100644 --- a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml +++ b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml @@ -41,11 +41,12 @@ Optimizer: no_weight_decay_name: absolute_pos_embed relative_position_bias_table .bias norm one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 20 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml index 96a9befd2026b70354dc20682d143c72b5660169..564da72f1fd4dc88b7161d00259a346a25b38c42 100644 --- a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml +++ b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml @@ -41,11 +41,12 @@ Optimizer: no_weight_decay_name: absolute_pos_embed relative_position_bias_table .bias norm one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 20 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml index ffbbcf080b025527035fee353c4ddc6a212c294e..ba42f1efb8460581445e7b5a605971ec64bb0851 100644 --- a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml +++ b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml @@ -41,11 +41,12 @@ Optimizer: no_weight_decay_name: absolute_pos_embed relative_position_bias_table .bias norm one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 20 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml index 066db715da45de4eb85ff4f3c2406be50be99e7e..26fa0ba61ed159ad458f9b0c21e03aa4fcd7f02e 100644 --- a/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml +++ b/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml @@ -41,11 +41,12 @@ Optimizer: no_weight_decay_name: absolute_pos_embed relative_position_bias_table .bias norm one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 20 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml b/ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml index 74c402ee7cdddba2234ec14f9388f83392c465e0..36e5b086dc43376fe6424ca67eb02b93ac6ce9a4 100644 --- a/ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml +++ b/ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: norm cls_token proj.0.weight proj.1.weight proj.2.weight proj.3.weight pos_block one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml b/ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml index ca66e9a33e81c87c7e119094069e36fed44bd608..6e19d64618fa6d1dbb5608558592760f6ad61cb7 100644 --- a/ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml +++ b/ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: norm cls_token proj.0.weight proj.1.weight proj.2.weight proj.3.weight pos_block one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml b/ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml index 9e97c0f9994da983a5951d784850096498549ced..66235960a5c6e3fb4c255da86031214856ea761f 100644 --- a/ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml +++ b/ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: norm cls_token proj.0.weight proj.1.weight proj.2.weight proj.3.weight pos_block one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/Twins/pcpvt_base.yaml b/ppcls/configs/ImageNet/Twins/pcpvt_base.yaml index 7831e9289570a62a60dd72945b2bdf842f9d8f09..96745495a0926bc2767a309e0c3bd71cf4201c0b 100644 --- a/ppcls/configs/ImageNet/Twins/pcpvt_base.yaml +++ b/ppcls/configs/ImageNet/Twins/pcpvt_base.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: norm cls_token proj.0.weight proj.1.weight proj.2.weight proj.3.weight pos_block one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/Twins/pcpvt_large.yaml b/ppcls/configs/ImageNet/Twins/pcpvt_large.yaml index 8e160b3c2100a7a6c2666df89e08a67e93d589ad..ca4baf942a4c0763ee03f4631030a9d2a0752e1d 100644 --- a/ppcls/configs/ImageNet/Twins/pcpvt_large.yaml +++ b/ppcls/configs/ImageNet/Twins/pcpvt_large.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: norm cls_token proj.0.weight proj.1.weight proj.2.weight proj.3.weight pos_block one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/Twins/pcpvt_small.yaml b/ppcls/configs/ImageNet/Twins/pcpvt_small.yaml index 582382d4d0fa38c89db337b0cc7acbcb79a9a544..a5e5f7e0564c9d6c96b51cbda0ff7312a899d6b0 100644 --- a/ppcls/configs/ImageNet/Twins/pcpvt_small.yaml +++ b/ppcls/configs/ImageNet/Twins/pcpvt_small.yaml @@ -43,11 +43,12 @@ Optimizer: no_weight_decay_name: norm cls_token proj.0.weight proj.1.weight proj.2.weight proj.3.weight pos_block one_dim_param_no_weight_decay: True lr: + # for 8 cards name: Cosine - learning_rate: 5e-4 - eta_min: 1e-5 + learning_rate: 1e-3 + eta_min: 2e-5 warmup_epoch: 5 - warmup_start_lr: 1e-6 + warmup_start_lr: 2e-6 # data loader for train and eval diff --git a/ppcls/configs/ImageNet/VAN/VAN_tiny.yaml b/ppcls/configs/ImageNet/VAN/VAN_tiny.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ae08dcfa9e99d0ccab4f051abe9351ff61c9b736 --- /dev/null +++ b/ppcls/configs/ImageNet/VAN/VAN_tiny.yaml @@ -0,0 +1,158 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: ./output/ + device: gpu + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 300 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 224, 224] + save_inference_dir: ./inference + # training model under @to_static + to_static: False + +# model architecture +Arch: + name: VAN_tiny + class_num: 1000 + drop_path_rate: 0.1 + drop_rate: 0.0 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: AdamW + beta1: 0.9 + beta2: 0.999 + epsilon: 1e-8 + weight_decay: 0.05 + one_dim_param_no_weight_decay: True + lr: + name: Cosine + learning_rate: 1e-3 + eta_min: 1e-6 + warmup_epoch: 5 + warmup_start_lr: 1e-6 + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/train_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - RandCropImage: + size: 224 + interpolation: random + backend: pil + - RandFlipImage: + flip_code: 1 + - TimmAutoAugment: + config_str: rand-m9-mstd0.5-inc1 + interpolation: random + img_size: 224 + mean: [0.5, 0.5, 0.5] + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.5, 0.5, 0.5] + std: [0.5, 0.5, 0.5] + order: '' + - RandomErasing: + EPSILON: 0.25 + sl: 0.02 + sh: 1.0/3.0 + r1: 0.3 + attempt: 10 + use_log_aspect: True + mode: pixel + batch_transform_ops: + - OpSampler: + MixupOperator: + alpha: 0.8 + prob: 0.5 + CutmixOperator: + alpha: 1.0 + prob: 0.5 + sampler: + name: DistributedBatchSampler + batch_size: 256 + drop_last: True + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + + Eval: + dataset: + name: ImageNetDataset + image_root: ./dataset/ILSVRC2012/ + cls_label_path: ./dataset/ILSVRC2012/val_list.txt + transform_ops: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.5, 0.5, 0.5] + std: [0.5, 0.5, 0.5] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 256 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Infer: + infer_imgs: docs/images/inference_deployment/whl_demo.jpg + batch_size: 10 + transforms: + - DecodeImage: + to_rgb: True + channel_first: False + - ResizeImage: + resize_short: 248 + interpolation: bicubic + backend: pil + - CropImage: + size: 224 + - NormalizeImage: + scale: 1.0/255.0 + mean: [0.5, 0.5, 0.5] + std: [0.5, 0.5, 0.5] + order: '' + - ToCHWImage: + PostProcess: + name: Topk + topk: 5 + class_id_map_file: ppcls/utils/imagenet1k_label_list.txt + +Metric: + Eval: + - TopkAcc: + topk: [1, 5] diff --git a/ppcls/configs/ResNet50_UReID_infer.yaml b/ppcls/configs/ResNet50_UReID_infer.yaml new file mode 100644 index 0000000000000000000000000000000000000000..750851d7cbaf7f11d97d78331368e20b82b9b7f6 --- /dev/null +++ b/ppcls/configs/ResNet50_UReID_infer.yaml @@ -0,0 +1,152 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + # pretrained_model: "./pd_model_trace/ISE/ISE_M_model" # pretrained ISE model for Market1501 + # pretrained_model: "./pd_model_trace/ISE/ISE_MS_model" # pretrained ISE model for MSMT17 + output_dir: "./output/" + device: "gpu" + save_interval: 1 + eval_during_train: True + eval_interval: 1 + epochs: 120 + print_batch_step: 10 + use_visualdl: False + # used for static mode and model export + image_shape: [3, 128, 256] + save_inference_dir: "./inference" + eval_mode: "retrieval" + +# model architecture +Arch: + name: "RecModel" + infer_output_key: "features" + infer_add_softmax: False + Backbone: + name: "ResNet50_last_stage_stride1" + pretrained: True + BackboneStopLayer: + name: "avg_pool" + Neck: + name: "BNNeck" + num_features: 2048 + Head: + name: "FC" + embedding_size: 2048 + class_num: 751 + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + - SupConLoss: + weight: 1.0 + views: 2 + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Momentum + momentum: 0.9 + lr: + name: Cosine + learning_rate: 0.04 + regularizer: + name: 'L2' + coeff: 0.0005 + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: "Market1501" # ["Market1501", "MSMT17"] + image_root: "./dataset" + cls_label_path: "bounding_box_train" + transform_ops: + - ResizeImage: + size: [128, 256] + interpolation: 'bicubic' + backend: 'pil' + - RandFlipImage: + flip_code: 1 + - Pad: + padding: 10 + fill: 0 + - RandomCrop: + size: [128, 256] + pad_if_needed: False + - NormalizeImage: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + - RandomErasing: + EPSILON: 0.5 + sl: 0.02 + sh: 0.4 + r1: 0.3 + mean: [0.485, 0.456, 0.406] + + sampler: + name: PKSampler + batch_size: 16 + sample_per_id: 4 + drop_last: True + shuffle: True + loader: + num_workers: 6 + use_shared_memory: True + Eval: + Query: + dataset: + name: "Market1501" # ["Market1501", "MSMT17"] + image_root: "./dataset" + cls_label_path: "query" + transform_ops: + - ResizeImage: + size: [128, 256] + interpolation: 'bicubic' + backend: 'pil' + - NormalizeImage: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 6 + use_shared_memory: True + + Gallery: + dataset: + name: "Market1501" # ["Market1501", "MSMT17"] + image_root: "./dataset" + cls_label_path: "bounding_box_test" + transform_ops: + - ResizeImage: + size: [128, 256] + interpolation: 'bicubic' + backend: 'pil' + - NormalizeImage: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + order: '' + sampler: + name: DistributedBatchSampler + batch_size: 64 + drop_last: False + shuffle: False + loader: + num_workers: 6 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] + - mAP: {} + diff --git a/ppcls/configs/reid/strong_baseline/baseline.yaml b/ppcls/configs/reid/strong_baseline/baseline.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d49deeeb4d27f304ac8de76b5feb6d314ed18a48 --- /dev/null +++ b/ppcls/configs/reid/strong_baseline/baseline.yaml @@ -0,0 +1,157 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: "./output/" + device: "gpu" + save_interval: 40 + eval_during_train: True + eval_interval: 10 + epochs: 120 + print_batch_step: 20 + use_visualdl: False + eval_mode: "retrieval" + retrieval_feature_from: "backbone" # 'backbone' or 'neck' + # used for static mode and model export + image_shape: [3, 256, 128] + save_inference_dir: "./inference" + +# model architecture +Arch: + name: "RecModel" + infer_output_key: "features" + infer_add_softmax: False + Backbone: + name: "ResNet50" + pretrained: True + stem_act: null + BackboneStopLayer: + name: "flatten" + Head: + name: "FC" + embedding_size: 2048 + class_num: 751 + weight_attr: + initializer: + name: KaimingUniform + fan_in: 12288 # 6*embedding_size + bias_attr: + initializer: + name: KaimingUniform + fan_in: 12288 # 6*embedding_size + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + - TripletLossV2: + weight: 1.0 + margin: 0.3 + normalize_feature: False + feature_from: "backbone" + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Adam + lr: + name: Piecewise + decay_epochs: [40, 70] + values: [0.00035, 0.000035, 0.0000035] + by_epoch: True + last_epoch: 0 + regularizer: + name: 'L2' + coeff: 0.0005 + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "bounding_box_train" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - RandFlipImage: + flip_code: 1 + - Pad: + padding: 10 + - RandCropImageV2: + size: [128, 256] + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + sampler: + name: DistributedRandomIdentitySampler + batch_size: 64 + num_instances: 4 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + Eval: + Query: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "query" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "bounding_box_test" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] + - mAP: {} diff --git a/ppcls/configs/reid/strong_baseline/softmax_triplet.yaml b/ppcls/configs/reid/strong_baseline/softmax_triplet.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c4d52cd1ef24745cd1d5fac7f7dec26fd98609e0 --- /dev/null +++ b/ppcls/configs/reid/strong_baseline/softmax_triplet.yaml @@ -0,0 +1,175 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: "./output/" + device: "gpu" + save_interval: 40 + eval_during_train: True + eval_interval: 10 + epochs: 120 + print_batch_step: 20 + use_visualdl: False + eval_mode: "retrieval" + retrieval_feature_from: "features" # 'backbone' or 'features' + # used for static mode and model export + image_shape: [3, 256, 128] + save_inference_dir: "./inference" + +# model architecture +Arch: + name: "RecModel" + infer_output_key: "features" + infer_add_softmax: False + Backbone: + name: "ResNet50_last_stage_stride1" + pretrained: True + stem_act: null + BackboneStopLayer: + name: "flatten" + Neck: + name: BNNeck + num_features: &feat_dim 2048 + weight_attr: + initializer: + name: Constant + value: 1.0 + bias_attr: + initializer: + name: Constant + value: 0.0 + learning_rate: 1.0e-20 # NOTE: Temporarily set lr small enough to freeze the bias to zero + Head: + name: "FC" + embedding_size: *feat_dim + class_num: 751 + weight_attr: + initializer: + name: Normal + std: 0.001 + bias_attr: False + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + - TripletLossV2: + weight: 1.0 + margin: 0.3 + normalize_feature: False + feature_from: "backbone" + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + name: Adam + lr: + name: Piecewise + decay_epochs: [30, 60] + values: [0.00035, 0.000035, 0.0000035] + warmup_epoch: 10 + warmup_start_lr: 0.0000035 + by_epoch: True + last_epoch: 0 + regularizer: + name: 'L2' + coeff: 0.0005 + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "bounding_box_train" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - RandFlipImage: + flip_code: 1 + - Pad: + padding: 10 + - RandCropImageV2: + size: [128, 256] + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - RandomErasing: + EPSILON: 0.5 + sl: 0.02 + sh: 0.4 + r1: 0.3 + mean: [0.485, 0.456, 0.406] + sampler: + name: DistributedRandomIdentitySampler + batch_size: 64 + num_instances: 4 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + Eval: + Query: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "query" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "bounding_box_test" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] + - mAP: {} diff --git a/ppcls/configs/reid/strong_baseline/softmax_triplet_with_center.yaml b/ppcls/configs/reid/strong_baseline/softmax_triplet_with_center.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2d14c3f9ce83f231c5a664f8d24177b25543b421 --- /dev/null +++ b/ppcls/configs/reid/strong_baseline/softmax_triplet_with_center.yaml @@ -0,0 +1,186 @@ +# global configs +Global: + checkpoints: null + pretrained_model: null + output_dir: "./output/" + device: "gpu" + save_interval: 40 + eval_during_train: True + eval_interval: 10 + epochs: 120 + print_batch_step: 20 + use_visualdl: False + eval_mode: "retrieval" + retrieval_feature_from: "features" # 'backbone' or 'features' + # used for static mode and model export + image_shape: [3, 256, 128] + save_inference_dir: "./inference" + +# model architecture +Arch: + name: "RecModel" + infer_output_key: "features" + infer_add_softmax: False + Backbone: + name: "ResNet50_last_stage_stride1" + pretrained: True + stem_act: null + BackboneStopLayer: + name: "flatten" + Neck: + name: BNNeck + num_features: &feat_dim 2048 + weight_attr: + initializer: + name: Constant + value: 1.0 + bias_attr: + initializer: + name: Constant + value: 0.0 + learning_rate: 1.0e-20 # NOTE: Temporarily set lr small enough to freeze the bias to zero + Head: + name: "FC" + embedding_size: *feat_dim + class_num: &class_num 751 + weight_attr: + initializer: + name: Normal + std: 0.001 + bias_attr: False + +# loss function config for traing/eval process +Loss: + Train: + - CELoss: + weight: 1.0 + epsilon: 0.1 + - TripletLossV2: + weight: 1.0 + margin: 0.3 + normalize_feature: False + feature_from: "backbone" + - CenterLoss: + weight: 0.0005 + num_classes: *class_num + feat_dim: *feat_dim + feature_from: "backbone" + Eval: + - CELoss: + weight: 1.0 + +Optimizer: + - Adam: + scope: RecModel + lr: + name: Piecewise + decay_epochs: [30, 60] + values: [0.00035, 0.000035, 0.0000035] + warmup_epoch: 10 + warmup_start_lr: 0.0000035 + by_epoch: True + last_epoch: 0 + regularizer: + name: 'L2' + coeff: 0.0005 + - SGD: + scope: CenterLoss + lr: + name: Constant + learning_rate: 1000.0 # NOTE: set to ori_lr*(1/centerloss_weight) to avoid manually scaling centers' gradidents. + +# data loader for train and eval +DataLoader: + Train: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "bounding_box_train" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - RandFlipImage: + flip_code: 1 + - Pad: + padding: 10 + - RandCropImageV2: + size: [128, 256] + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + - RandomErasing: + EPSILON: 0.5 + sl: 0.02 + sh: 0.4 + r1: 0.3 + mean: [0.485, 0.456, 0.406] + sampler: + name: DistributedRandomIdentitySampler + batch_size: 64 + num_instances: 4 + drop_last: False + shuffle: True + loader: + num_workers: 4 + use_shared_memory: True + Eval: + Query: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "query" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + + Gallery: + dataset: + name: "Market1501" + image_root: "./dataset/" + cls_label_path: "bounding_box_test" + backend: "pil" + transform_ops: + - ResizeImage: + size: [128, 256] + return_numpy: False + interpolation: 'bilinear' + backend: "pil" + - ToTensor: + - Normalize: + mean: [0.485, 0.456, 0.406] + std: [0.229, 0.224, 0.225] + sampler: + name: DistributedBatchSampler + batch_size: 128 + drop_last: False + shuffle: False + loader: + num_workers: 4 + use_shared_memory: True + +Metric: + Eval: + - Recallk: + topk: [1, 5] + - mAP: {} diff --git a/ppcls/data/__init__.py b/ppcls/data/__init__.py index c763e454c24448174e4d601a46509f1bc40dc786..9fc4d760be545ffa93652c80d285e17ad0c8ae57 100644 --- a/ppcls/data/__init__.py +++ b/ppcls/data/__init__.py @@ -29,6 +29,8 @@ from ppcls.data.dataloader.logo_dataset import LogoDataset from ppcls.data.dataloader.icartoon_dataset import ICartoonDataset from ppcls.data.dataloader.mix_dataset import MixDataset from ppcls.data.dataloader.multi_scale_dataset import MultiScaleDataset +from ppcls.data.dataloader.person_dataset import Market1501, MSMT17 + # sampler from ppcls.data.dataloader.DistributedRandomIdentitySampler import DistributedRandomIdentitySampler diff --git a/ppcls/data/dataloader/__init__.py b/ppcls/data/dataloader/__init__.py index d9dcaeb2fe3e48892ccd1c190f372d5c832ffe57..2b1d92b76bd202e36086f21a3a092c3673277690 100644 --- a/ppcls/data/dataloader/__init__.py +++ b/ppcls/data/dataloader/__init__.py @@ -9,3 +9,4 @@ from ppcls.data.dataloader.multi_scale_dataset import MultiScaleDataset from ppcls.data.dataloader.mix_sampler import MixSampler from ppcls.data.dataloader.multi_scale_sampler import MultiScaleSampler from ppcls.data.dataloader.pk_sampler import PKSampler +from ppcls.data.dataloader.person_dataset import Market1501, MSMT17 diff --git a/ppcls/data/dataloader/dali.py b/ppcls/data/dataloader/dali.py index a15c231568a97fd607f2ada4f5f6e81fa084cc62..a340a946c921bedd475531eb3bd9172f49a99e1e 100644 --- a/ppcls/data/dataloader/dali.py +++ b/ppcls/data/dataloader/dali.py @@ -230,7 +230,7 @@ def dali_dataloader(config, mode, device, seed=None): lower = ratio[0] upper = ratio[1] - if 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env: + if 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env and 'FLAGS_selected_gpus' in env: shard_id = int(env['PADDLE_TRAINER_ID']) num_shards = int(env['PADDLE_TRAINERS_NUM']) device_id = int(env['FLAGS_selected_gpus']) @@ -282,7 +282,7 @@ def dali_dataloader(config, mode, device, seed=None): else: resize_shorter = transforms["ResizeImage"].get("resize_short", 256) crop = transforms["CropImage"]["size"] - if 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env and sampler_name == "DistributedBatchSampler": + if 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env and 'FLAGS_selected_gpus' in env and sampler_name == "DistributedBatchSampler": shard_id = int(env['PADDLE_TRAINER_ID']) num_shards = int(env['PADDLE_TRAINERS_NUM']) device_id = int(env['FLAGS_selected_gpus']) diff --git a/ppcls/data/dataloader/imagenet_dataset.py b/ppcls/data/dataloader/imagenet_dataset.py index 1166ab3851b0c594469f135c1c4b3c7bc921ac5f..0e2b1c8a0b4cbc7e2f251ed2363fca71ec3239e7 100644 --- a/ppcls/data/dataloader/imagenet_dataset.py +++ b/ppcls/data/dataloader/imagenet_dataset.py @@ -21,6 +21,15 @@ from .common_dataset import CommonDataset class ImageNetDataset(CommonDataset): + def __init__( + self, + image_root, + cls_label_path, + transform_ops=None, + delimiter=None): + self.delimiter = delimiter if delimiter is not None else " " + super(ImageNetDataset, self).__init__(image_root, cls_label_path, transform_ops) + def _load_anno(self, seed=None): assert os.path.exists(self._cls_path) assert os.path.exists(self._img_root) @@ -32,7 +41,7 @@ class ImageNetDataset(CommonDataset): if seed is not None: np.random.RandomState(seed).shuffle(lines) for l in lines: - l = l.strip().split(" ") + l = l.strip().split(self.delimiter) self.images.append(os.path.join(self._img_root, l[0])) self.labels.append(np.int64(l[1])) assert os.path.exists(self.images[-1]) diff --git a/ppcls/data/dataloader/person_dataset.py b/ppcls/data/dataloader/person_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..97af957c4aef31d7a1b691f2a5f5c037d07deea4 --- /dev/null +++ b/ppcls/data/dataloader/person_dataset.py @@ -0,0 +1,224 @@ +# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import print_function + +import numpy as np +import paddle +from paddle.io import Dataset +import os +import cv2 + +from ppcls.data import preprocess +from ppcls.data.preprocess import transform +from ppcls.utils import logger +from .common_dataset import create_operators +import os.path as osp +import glob +import re +from PIL import Image + + +class Market1501(Dataset): + """ + Market1501 + Reference: + Zheng et al. Scalable Person Re-identification: A Benchmark. ICCV 2015. + URL: http://www.liangzheng.org/Project/project_reid.html + + Dataset statistics: + # identities: 1501 (+1 for background) + # images: 12936 (train) + 3368 (query) + 15913 (gallery) + """ + _dataset_dir = 'market1501/Market-1501-v15.09.15' + + def __init__(self, + image_root, + cls_label_path, + transform_ops=None, + backend="cv2"): + self._img_root = image_root + self._cls_path = cls_label_path # the sub folder in the dataset + self._dataset_dir = osp.join(image_root, self._dataset_dir, + self._cls_path) + self._check_before_run() + if transform_ops: + self._transform_ops = create_operators(transform_ops) + self.backend = backend + self._dtype = paddle.get_default_dtype() + self._load_anno(relabel=True if 'train' in self._cls_path else False) + + def _check_before_run(self): + """Check if the file is available before going deeper""" + if not osp.exists(self._dataset_dir): + raise RuntimeError("'{}' is not available".format( + self._dataset_dir)) + + def _load_anno(self, relabel=False): + img_paths = glob.glob(osp.join(self._dataset_dir, '*.jpg')) + pattern = re.compile(r'([-\d]+)_c(\d)') + + self.images = [] + self.labels = [] + self.cameras = [] + pid_container = set() + + for img_path in sorted(img_paths): + pid, _ = map(int, pattern.search(img_path).groups()) + if pid == -1: continue # junk images are just ignored + pid_container.add(pid) + pid2label = {pid: label for label, pid in enumerate(pid_container)} + + for img_path in sorted(img_paths): + pid, camid = map(int, pattern.search(img_path).groups()) + if pid == -1: continue # junk images are just ignored + assert 0 <= pid <= 1501 # pid == 0 means background + assert 1 <= camid <= 6 + camid -= 1 # index starts from 0 + if relabel: pid = pid2label[pid] + self.images.append(img_path) + self.labels.append(pid) + self.cameras.append(camid) + + self.num_pids, self.num_imgs, self.num_cams = get_imagedata_info( + self.images, self.labels, self.cameras, subfolder=self._cls_path) + + def __getitem__(self, idx): + try: + img = Image.open(self.images[idx]).convert('RGB') + if self.backend == "cv2": + img = np.array(img, dtype="float32").astype(np.uint8) + if self._transform_ops: + img = transform(img, self._transform_ops) + if self.backend == "cv2": + img = img.transpose((2, 0, 1)) + return (img, self.labels[idx], self.cameras[idx]) + except Exception as ex: + logger.error("Exception occured when parse line: {} with msg: {}". + format(self.images[idx], ex)) + rnd_idx = np.random.randint(self.__len__()) + return self.__getitem__(rnd_idx) + + def __len__(self): + return len(self.images) + + @property + def class_num(self): + return len(set(self.labels)) + + +class MSMT17(Dataset): + """ + MSMT17 + + Reference: + Wei et al. Person Transfer GAN to Bridge Domain Gap for Person Re-Identification. CVPR 2018. + + URL: http://www.pkuvmc.com/publications/msmt17.html + + Dataset statistics: + # identities: 4101 + # images: 32621 (train) + 11659 (query) + 82161 (gallery) + # cameras: 15 + """ + _dataset_dir = 'msmt17/MSMT17_V1' + + def __init__(self, image_root, cls_label_path, transform_ops=None): + self._img_root = image_root + self._cls_path = cls_label_path # the sub folder in the dataset + self._dataset_dir = osp.join(image_root, self._dataset_dir, + self._cls_path) + self._check_before_run() + if transform_ops: + self._transform_ops = create_operators(transform_ops) + self._dtype = paddle.get_default_dtype() + self._load_anno(relabel=True if 'train' in self._cls_path else False) + + def _check_before_run(self): + """Check if the file is available before going deeper""" + if not osp.exists(self._dataset_dir): + raise RuntimeError("'{}' is not available".format( + self._dataset_dir)) + + def _load_anno(self, relabel=False): + img_paths = glob.glob(osp.join(self._dataset_dir, '*.jpg')) + pattern = re.compile(r'([-\d]+)_c(\d+)') + + self.images = [] + self.labels = [] + self.cameras = [] + pid_container = set() + + for img_path in img_paths: + pid, _ = map(int, pattern.search(img_path).groups()) + if pid == -1: + continue # junk images are just ignored + pid_container.add(pid) + pid2label = {pid: label for label, pid in enumerate(pid_container)} + + for img_path in img_paths: + pid, camid = map(int, pattern.search(img_path).groups()) + if pid == -1: + continue # junk images are just ignored + assert 1 <= camid <= 15 + camid -= 1 # index starts from 0 + if relabel: + pid = pid2label[pid] + self.images.append(img_path) + self.labels.append(pid) + self.cameras.append(camid) + + self.num_pids, self.num_imgs, self.num_cams = get_imagedata_info( + self.images, self.labels, self.cameras, subfolder=self._cls_path) + + def __getitem__(self, idx): + try: + img = Image.open(self.images[idx]).convert('RGB') + img = np.array(img, dtype="float32").astype(np.uint8) + if self._transform_ops: + img = transform(img, self._transform_ops) + img = img.transpose((2, 0, 1)) + return (img, self.labels[idx], self.cameras[idx]) + except Exception as ex: + logger.error("Exception occured when parse line: {} with msg: {}". + format(self.images[idx], ex)) + rnd_idx = np.random.randint(self.__len__()) + return self.__getitem__(rnd_idx) + + def __len__(self): + return len(self.images) + + @property + def class_num(self): + return len(set(self.labels)) + + +def get_imagedata_info(data, labels, cameras, subfolder='train'): + pids, cams = [], [] + for _, pid, camid in zip(data, labels, cameras): + pids += [pid] + cams += [camid] + pids = set(pids) + cams = set(cams) + num_pids = len(pids) + num_cams = len(cams) + num_imgs = len(data) + print("Dataset statistics:") + print(" ----------------------------------------") + print(" subset | # ids | # images | # cameras") + print(" ----------------------------------------") + print(" {} | {:5d} | {:8d} | {:9d}".format(subfolder, num_pids, + num_imgs, num_cams)) + print(" ----------------------------------------") + return num_pids, num_imgs, num_cams diff --git a/ppcls/data/postprocess/topk.py b/ppcls/data/postprocess/topk.py index 9c1371bfd11f4c93f06c82436e88e0ff20a57b35..df02719471300ea8e2b7c1db286d104adabe116f 100644 --- a/ppcls/data/postprocess/topk.py +++ b/ppcls/data/postprocess/topk.py @@ -19,10 +19,11 @@ import paddle.nn.functional as F class Topk(object): - def __init__(self, topk=1, class_id_map_file=None): + def __init__(self, topk=1, class_id_map_file=None, delimiter=None): assert isinstance(topk, (int, )) self.class_id_map = self.parse_class_id_map(class_id_map_file) self.topk = topk + self.delimiter = delimiter if delimiter is not None else " " def parse_class_id_map(self, class_id_map_file): if class_id_map_file is None: @@ -38,7 +39,7 @@ class Topk(object): with open(class_id_map_file, "r") as fin: lines = fin.readlines() for line in lines: - partition = line.split("\n")[0].partition(" ") + partition = line.split("\n")[0].partition(self.delimiter) class_id_map[int(partition[0])] = str(partition[-1]) except Exception as ex: print(ex) @@ -46,6 +47,8 @@ class Topk(object): return class_id_map def __call__(self, x, file_names=None, multilabel=False): + if isinstance(x, dict): + x = x['logits'] assert isinstance(x, paddle.Tensor) if file_names is not None: assert x.shape[0] == len(file_names) diff --git a/ppcls/data/preprocess/__init__.py b/ppcls/data/preprocess/__init__.py index 075ee89278e2e099ce3c9cbc108dfe159e2012f2..62066016a47c8cef7bd31bc7d238f202ea6455f0 100644 --- a/ppcls/data/preprocess/__init__.py +++ b/ppcls/data/preprocess/__init__.py @@ -25,10 +25,14 @@ from ppcls.data.preprocess.ops.operators import DecodeImage from ppcls.data.preprocess.ops.operators import ResizeImage from ppcls.data.preprocess.ops.operators import CropImage from ppcls.data.preprocess.ops.operators import RandCropImage +from ppcls.data.preprocess.ops.operators import RandCropImageV2 from ppcls.data.preprocess.ops.operators import RandFlipImage from ppcls.data.preprocess.ops.operators import NormalizeImage from ppcls.data.preprocess.ops.operators import ToCHWImage from ppcls.data.preprocess.ops.operators import AugMix +from ppcls.data.preprocess.ops.operators import Pad +from ppcls.data.preprocess.ops.operators import ToTensor +from ppcls.data.preprocess.ops.operators import Normalize from ppcls.data.preprocess.batch_ops.batch_operators import MixupOperator, CutmixOperator, OpSampler, FmixOperator diff --git a/ppcls/data/preprocess/ops/autoaugment.py b/ppcls/data/preprocess/ops/autoaugment.py index 330220a93c8b304c38603644582a5f5efd623b03..dba27808c79ea00d02e98e33d9739e2cbf8ffb04 100644 --- a/ppcls/data/preprocess/ops/autoaugment.py +++ b/ppcls/data/preprocess/ops/autoaugment.py @@ -13,6 +13,7 @@ # limitations under the License. # This code is based on https://github.com/DeepVoltaire/AutoAugment/blob/master/autoaugment.py +# reference: https://arxiv.org/abs/1805.09501 from PIL import Image, ImageEnhance, ImageOps import numpy as np diff --git a/ppcls/data/preprocess/ops/cutout.py b/ppcls/data/preprocess/ops/cutout.py index b906e14520429b42c800ec3eb676fcc8587eb7da..d442fd138f1ca5d803dceeb133cf9b61a39d87cb 100644 --- a/ppcls/data/preprocess/ops/cutout.py +++ b/ppcls/data/preprocess/ops/cutout.py @@ -13,6 +13,7 @@ # limitations under the License. # This code is based on https://github.com/uoguelph-mlrg/Cutout +# reference: https://arxiv.org/abs/1708.04552 import numpy as np import random diff --git a/ppcls/data/preprocess/ops/fmix.py b/ppcls/data/preprocess/ops/fmix.py index dc2ef9120a7fd5858d24d9acff9bf741623e731b..019f618c57400a6d51fd3eed94c23fc1695c383b 100644 --- a/ppcls/data/preprocess/ops/fmix.py +++ b/ppcls/data/preprocess/ops/fmix.py @@ -12,6 +12,9 @@ # See the License for the specific language governing permissions and # limitations under the License. +# This code is based on https://github.com/ecs-vlc/FMix +# reference: https://arxiv.org/abs/2002.12047 + import math import random diff --git a/ppcls/data/preprocess/ops/grid.py b/ppcls/data/preprocess/ops/grid.py index 6f0b2dc8dce3953a7f7595621acf1d4cde447651..1a9a76d86f6d050b4519d7527de821bfd70cede2 100644 --- a/ppcls/data/preprocess/ops/grid.py +++ b/ppcls/data/preprocess/ops/grid.py @@ -13,6 +13,7 @@ # limitations under the License. # This code is based on https://github.com/akuxcw/GridMask +# reference: https://arxiv.org/abs/2001.04086. import numpy as np from PIL import Image diff --git a/ppcls/data/preprocess/ops/hide_and_seek.py b/ppcls/data/preprocess/ops/hide_and_seek.py index 33f25f26552b5030cc69531fcaf4d3705dc05c87..16fc671cf65958d23e720c07d9595022cfe3756c 100644 --- a/ppcls/data/preprocess/ops/hide_and_seek.py +++ b/ppcls/data/preprocess/ops/hide_and_seek.py @@ -13,6 +13,7 @@ # limitations under the License. # This code is based on https://github.com/kkanshul/Hide-and-Seek +# reference: http://krsingh.cs.ucdavis.edu/krishna_files/papers/hide_and_seek/my_files/iccv2017.pdf import numpy as np import random diff --git a/ppcls/data/preprocess/ops/operators.py b/ppcls/data/preprocess/ops/operators.py index 9cdc58b2b5f6c4987c898a6549f5652797d1c845..157f44f1ab15ffd1162aeada37dba9296ee0ca00 100644 --- a/ppcls/data/preprocess/ops/operators.py +++ b/ppcls/data/preprocess/ops/operators.py @@ -23,8 +23,9 @@ import math import random import cv2 import numpy as np -from PIL import Image +from PIL import Image, ImageOps, __version__ as PILLOW_VERSION from paddle.vision.transforms import ColorJitter as RawColorJitter +from paddle.vision.transforms import ToTensor, Normalize from .autoaugment import ImageNetPolicy from .functional import augmentations @@ -32,13 +33,14 @@ from ppcls.utils import logger class UnifiedResize(object): - def __init__(self, interpolation=None, backend="cv2"): + def __init__(self, interpolation=None, backend="cv2", return_numpy=True): _cv2_interp_from_str = { 'nearest': cv2.INTER_NEAREST, 'bilinear': cv2.INTER_LINEAR, 'area': cv2.INTER_AREA, 'bicubic': cv2.INTER_CUBIC, - 'lanczos': cv2.INTER_LANCZOS4 + 'lanczos': cv2.INTER_LANCZOS4, + 'random': (cv2.INTER_LINEAR, cv2.INTER_CUBIC) } _pil_interp_from_str = { 'nearest': Image.NEAREST, @@ -46,13 +48,26 @@ class UnifiedResize(object): 'bicubic': Image.BICUBIC, 'box': Image.BOX, 'lanczos': Image.LANCZOS, - 'hamming': Image.HAMMING + 'hamming': Image.HAMMING, + 'random': (Image.BILINEAR, Image.BICUBIC) } - def _pil_resize(src, size, resample): - pil_img = Image.fromarray(src) + def _cv2_resize(src, size, resample): + if isinstance(resample, tuple): + resample = random.choice(resample) + return cv2.resize(src, size, interpolation=resample) + + def _pil_resize(src, size, resample, return_numpy=True): + if isinstance(resample, tuple): + resample = random.choice(resample) + if isinstance(src, np.ndarray): + pil_img = Image.fromarray(src) + else: + pil_img = src pil_img = pil_img.resize(size, resample) - return np.asarray(pil_img) + if return_numpy: + return np.asarray(pil_img) + return pil_img if backend.lower() == "cv2": if isinstance(interpolation, str): @@ -60,11 +75,12 @@ class UnifiedResize(object): # compatible with opencv < version 4.4.0 elif interpolation is None: interpolation = cv2.INTER_LINEAR - self.resize_func = partial(cv2.resize, interpolation=interpolation) + self.resize_func = partial(_cv2_resize, resample=interpolation) elif backend.lower() == "pil": if isinstance(interpolation, str): interpolation = _pil_interp_from_str[interpolation.lower()] - self.resize_func = partial(_pil_resize, resample=interpolation) + self.resize_func = partial( + _pil_resize, resample=interpolation, return_numpy=return_numpy) else: logger.warning( f"The backend of Resize only support \"cv2\" or \"PIL\". \"f{backend}\" is unavailable. Use \"cv2\" instead." @@ -72,6 +88,8 @@ class UnifiedResize(object): self.resize_func = cv2.resize def __call__(self, src, size): + if isinstance(size, list): + size = tuple(size) return self.resize_func(src, size) @@ -90,14 +108,15 @@ class DecodeImage(object): self.channel_first = channel_first # only enabled when to_np is True def __call__(self, img): - if six.PY2: - assert type(img) is str and len( - img) > 0, "invalid input 'img' in DecodeImage" - else: - assert type(img) is bytes and len( - img) > 0, "invalid input 'img' in DecodeImage" - data = np.frombuffer(img, dtype='uint8') - img = cv2.imdecode(data, 1) + if not isinstance(img, np.ndarray): + if six.PY2: + assert type(img) is str and len( + img) > 0, "invalid input 'img' in DecodeImage" + else: + assert type(img) is bytes and len( + img) > 0, "invalid input 'img' in DecodeImage" + data = np.frombuffer(img, dtype='uint8') + img = cv2.imdecode(data, 1) if self.to_rgb: assert img.shape[2] == 3, 'invalid shape of image[%s]' % ( img.shape) @@ -116,7 +135,8 @@ class ResizeImage(object): size=None, resize_short=None, interpolation=None, - backend="cv2"): + backend="cv2", + return_numpy=True): if resize_short is not None and resize_short > 0: self.resize_short = resize_short self.w = None @@ -130,10 +150,16 @@ class ResizeImage(object): 'both 'size' and 'resize_short' are None") self._resize_func = UnifiedResize( - interpolation=interpolation, backend=backend) + interpolation=interpolation, + backend=backend, + return_numpy=return_numpy) def __call__(self, img): - img_h, img_w = img.shape[:2] + if isinstance(img, np.ndarray): + img_h, img_w = img.shape[:2] + else: + img_w, img_h = img.size + if self.resize_short is not None: percent = float(self.resize_short) / min(img_w, img_h) w = int(round(img_w * percent)) @@ -213,6 +239,40 @@ class RandCropImage(object): return self._resize_func(img, size) +class RandCropImageV2(object): + """ RandCropImageV2 is different from RandCropImage, + it will Select a cutting position randomly in a uniform distribution way, + and cut according to the given size without resize at last.""" + + def __init__(self, size): + if type(size) is int: + self.size = (size, size) # (h, w) + else: + self.size = size + + def __call__(self, img): + if isinstance(img, np.ndarray): + img_h, img_w = img.shap[0], img.shap[1] + else: + img_w, img_h = img.size + tw, th = self.size + + if img_h + 1 < th or img_w + 1 < tw: + raise ValueError( + "Required crop size {} is larger then input image size {}". + format((th, tw), (img_h, img_w))) + + if img_w == tw and img_h == th: + return img + + top = random.randint(0, img_h - th + 1) + left = random.randint(0, img_w - tw + 1) + if isinstance(img, np.ndarray): + return img[top:top + th, left:left + tw, :] + else: + return img.crop((left, top, left + tw, top + th)) + + class RandFlipImage(object): """ random flip image flip_code: @@ -228,7 +288,10 @@ class RandFlipImage(object): def __call__(self, img): if random.randint(0, 1) == 1: - return cv2.flip(img, self.flip_code) + if isinstance(img, np.ndarray): + return cv2.flip(img, self.flip_code) + else: + return img.transpose(Image.FLIP_LEFT_RIGHT) else: return img @@ -382,3 +445,58 @@ class ColorJitter(RawColorJitter): if isinstance(img, Image.Image): img = np.asarray(img) return img + + +class Pad(object): + """ + Pads the given PIL.Image on all sides with specified padding mode and fill value. + adapted from: https://pytorch.org/vision/stable/_modules/torchvision/transforms/transforms.html#Pad + """ + + def __init__(self, padding: int, fill: int=0, + padding_mode: str="constant"): + self.padding = padding + self.fill = fill + self.padding_mode = padding_mode + + def _parse_fill(self, fill, img, min_pil_version, name="fillcolor"): + # Process fill color for affine transforms + major_found, minor_found = (int(v) + for v in PILLOW_VERSION.split('.')[:2]) + major_required, minor_required = ( + int(v) for v in min_pil_version.split('.')[:2]) + if major_found < major_required or (major_found == major_required and + minor_found < minor_required): + if fill is None: + return {} + else: + msg = ( + "The option to fill background area of the transformed image, " + "requires pillow>={}") + raise RuntimeError(msg.format(min_pil_version)) + + num_bands = len(img.getbands()) + if fill is None: + fill = 0 + if isinstance(fill, (int, float)) and num_bands > 1: + fill = tuple([fill] * num_bands) + if isinstance(fill, (list, tuple)): + if len(fill) != num_bands: + msg = ( + "The number of elements in 'fill' does not match the number of " + "bands of the image ({} != {})") + raise ValueError(msg.format(len(fill), num_bands)) + + fill = tuple(fill) + + return {name: fill} + + def __call__(self, img): + opts = self._parse_fill(self.fill, img, "2.3.0", name="fill") + if img.mode == "P": + palette = img.getpalette() + img = ImageOps.expand(img, border=self.padding, **opts) + img.putpalette(palette) + return img + + return ImageOps.expand(img, border=self.padding, **opts) diff --git a/ppcls/data/preprocess/ops/randaugment.py b/ppcls/data/preprocess/ops/randaugment.py index cca59da4236169f927deae6385b76aa1802f61f6..98df62d6b1154453702880f59c5d3079ff815d0b 100644 --- a/ppcls/data/preprocess/ops/randaugment.py +++ b/ppcls/data/preprocess/ops/randaugment.py @@ -13,6 +13,7 @@ # limitations under the License. # This code is based on https://github.com/heartInsert/randaugment +# reference: https://arxiv.org/abs/1909.13719 from PIL import Image, ImageEnhance, ImageOps import numpy as np diff --git a/ppcls/data/preprocess/ops/random_erasing.py b/ppcls/data/preprocess/ops/random_erasing.py index f234abbbac87cf8230e4d619fe7832e8309abcdb..648b41ea532eb8a767015de6abcdf7fc0448e34c 100644 --- a/ppcls/data/preprocess/ops/random_erasing.py +++ b/ppcls/data/preprocess/ops/random_erasing.py @@ -12,7 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. -#This code is adapted from https://github.com/zhunzhong07/Random-Erasing, and refer to Timm. +# This code is adapted from https://github.com/zhunzhong07/Random-Erasing, and refer to Timm(https://github.com/rwightman/pytorch-image-models). +# reference: https://arxiv.org/abs/1708.04896 from functools import partial @@ -25,15 +26,21 @@ import numpy as np class Pixels(object): def __init__(self, mode="const", mean=[0., 0., 0.]): self._mode = mode - self._mean = mean + self._mean = np.array(mean) - def __call__(self, h=224, w=224, c=3): + def __call__(self, h=224, w=224, c=3, channel_first=False): if self._mode == "rand": - return np.random.normal(size=(1, 1, 3)) + return np.random.normal(size=( + 1, 1, 3)) if not channel_first else np.random.normal(size=( + 3, 1, 1)) elif self._mode == "pixel": - return np.random.normal(size=(h, w, c)) + return np.random.normal(size=( + h, w, c)) if not channel_first else np.random.normal(size=( + c, h, w)) elif self._mode == "const": - return self._mean + return np.reshape(self._mean, ( + 1, 1, c)) if not channel_first else np.reshape(self._mean, + (c, 1, 1)) else: raise Exception( "Invalid mode in RandomErasing, only support \"const\", \"rand\", \"pixel\"" @@ -68,7 +75,13 @@ class RandomErasing(object): return img for _ in range(self.attempt): - area = img.shape[0] * img.shape[1] + if isinstance(img, np.ndarray): + img_h, img_w, img_c = img.shape + channel_first = False + else: + img_c, img_h, img_w = img.shape + channel_first = True + area = img_h * img_w target_area = random.uniform(self.sl, self.sh) * area aspect_ratio = random.uniform(*self.r1) @@ -78,13 +91,19 @@ class RandomErasing(object): h = int(round(math.sqrt(target_area * aspect_ratio))) w = int(round(math.sqrt(target_area / aspect_ratio))) - if w < img.shape[1] and h < img.shape[0]: - pixels = self.get_pixels(h, w, img.shape[2]) - x1 = random.randint(0, img.shape[0] - h) - y1 = random.randint(0, img.shape[1] - w) - if img.shape[2] == 3: - img[x1:x1 + h, y1:y1 + w, :] = pixels + if w < img_w and h < img_h: + pixels = self.get_pixels(h, w, img_c, channel_first) + x1 = random.randint(0, img_h - h) + y1 = random.randint(0, img_w - w) + if img_c == 3: + if channel_first: + img[:, x1:x1 + h, y1:y1 + w] = pixels + else: + img[x1:x1 + h, y1:y1 + w, :] = pixels else: - img[x1:x1 + h, y1:y1 + w, 0] = pixels[0] + if channel_first: + img[0, x1:x1 + h, y1:y1 + w] = pixels[0] + else: + img[x1:x1 + h, y1:y1 + w, 0] = pixels[:, :, 0] return img return img diff --git a/ppcls/data/preprocess/ops/timm_autoaugment.py b/ppcls/data/preprocess/ops/timm_autoaugment.py index dd2994dac9aed5be398f02bbc07ae66c176911ca..30f1f505ae512d6200951bd305b49a03fa6cb9b6 100644 --- a/ppcls/data/preprocess/ops/timm_autoaugment.py +++ b/ppcls/data/preprocess/ops/timm_autoaugment.py @@ -12,7 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. -# Code was heavily based on https://github.com/rwightman/pytorch-image-models +# This code is heavily based on https://github.com/rwightman/pytorch-image-models +# reference: https://arxiv.org/abs/1805.09501 import random import math diff --git a/ppcls/engine/engine.py b/ppcls/engine/engine.py index 7f04221c8664e9b8a7a7c3624d550c044409ddce..5b5c4da8a6500ab90c31f33097075db5f8ee5f89 100644 --- a/ppcls/engine/engine.py +++ b/ppcls/engine/engine.py @@ -71,7 +71,7 @@ class Engine(object): self.output_dir = self.config['Global']['output_dir'] log_file = os.path.join(self.output_dir, self.config["Arch"]["name"], f"{mode}.log") - init_logger(name='root', log_file=log_file) + init_logger(log_file=log_file) print_config(config) # init train_func and eval_func @@ -92,27 +92,12 @@ class Engine(object): self.vdl_writer = LogWriter(logdir=vdl_writer_path) # set device - assert self.config["Global"]["device"] in ["cpu", "gpu", "xpu", "npu"] + assert self.config["Global"][ + "device"] in ["cpu", "gpu", "xpu", "npu", "mlu"] self.device = paddle.set_device(self.config["Global"]["device"]) logger.info('train with paddle {} and device {}'.format( paddle.__version__, self.device)) - # AMP training - self.amp = True if "AMP" in self.config and self.mode == "train" else False - if self.amp and self.config["AMP"] is not None: - self.scale_loss = self.config["AMP"].get("scale_loss", 1.0) - self.use_dynamic_loss_scaling = self.config["AMP"].get( - "use_dynamic_loss_scaling", False) - else: - self.scale_loss = 1.0 - self.use_dynamic_loss_scaling = False - if self.amp: - AMP_RELATED_FLAGS_SETTING = { - 'FLAGS_cudnn_batchnorm_spatial_persistent': 1, - 'FLAGS_max_inplace_grad_add': 8, - } - paddle.fluid.set_flags(AMP_RELATED_FLAGS_SETTING) - if "class_num" in config["Global"]: global_class_num = config["Global"]["class_num"] if "class_num" not in config["Arch"]: @@ -171,7 +156,9 @@ class Engine(object): if metric_config is not None: metric_config = metric_config.get("Train") if metric_config is not None: - if hasattr(self.train_dataloader, "collate_fn"): + if hasattr( + self.train_dataloader, "collate_fn" + ) and self.train_dataloader.collate_fn is not None: for m_idx, m in enumerate(metric_config): if "TopkAcc" in m: msg = f"'TopkAcc' metric can not be used when setting 'batch_transform_ops' in config. The 'TopkAcc' metric has been removed." @@ -210,44 +197,109 @@ class Engine(object): if self.config["Global"]["pretrained_model"] is not None: if self.config["Global"]["pretrained_model"].startswith("http"): load_dygraph_pretrain_from_url( - self.model, self.config["Global"]["pretrained_model"]) + [self.model, getattr(self, 'train_loss_func', None)], + self.config["Global"]["pretrained_model"]) else: load_dygraph_pretrain( - self.model, self.config["Global"]["pretrained_model"]) + [self.model, getattr(self, 'train_loss_func', None)], + self.config["Global"]["pretrained_model"]) # build optimizer if self.mode == 'train': self.optimizer, self.lr_sch = build_optimizer( self.config["Optimizer"], self.config["Global"]["epochs"], - len(self.train_dataloader), [self.model]) + len(self.train_dataloader), + [self.model, self.train_loss_func]) - # for amp training + # AMP training and evaluating + self.amp = "AMP" in self.config and self.config["AMP"] is not None + self.amp_eval = False + # for amp if self.amp: + AMP_RELATED_FLAGS_SETTING = {'FLAGS_max_inplace_grad_add': 8, } + if paddle.is_compiled_with_cuda(): + AMP_RELATED_FLAGS_SETTING.update({ + 'FLAGS_cudnn_batchnorm_spatial_persistent': 1 + }) + paddle.fluid.set_flags(AMP_RELATED_FLAGS_SETTING) + + self.scale_loss = self.config["AMP"].get("scale_loss", 1.0) + self.use_dynamic_loss_scaling = self.config["AMP"].get( + "use_dynamic_loss_scaling", False) self.scaler = paddle.amp.GradScaler( init_loss_scaling=self.scale_loss, use_dynamic_loss_scaling=self.use_dynamic_loss_scaling) - amp_level = self.config['AMP'].get("level", "O1") - if amp_level not in ["O1", "O2"]: + + self.amp_level = self.config['AMP'].get("level", "O1") + if self.amp_level not in ["O1", "O2"]: msg = "[Parameter Error]: The optimize level of AMP only support 'O1' and 'O2'. The level has been set 'O1'." logger.warning(msg) self.config['AMP']["level"] = "O1" - amp_level = "O1" - self.model, self.optimizer = paddle.amp.decorate( - models=self.model, - optimizers=self.optimizer, - level=amp_level, - save_dtype='float32') - - # for distributed + self.amp_level = "O1" + + self.amp_eval = self.config["AMP"].get("use_fp16_test", False) + # TODO(gaotingquan): Paddle not yet support FP32 evaluation when training with AMPO2 + if self.config["Global"].get( + "eval_during_train", + True) and self.amp_level == "O2" and self.amp_eval == False: + msg = "PaddlePaddle only support FP16 evaluation when training with AMP O2 now. " + logger.warning(msg) + self.config["AMP"]["use_fp16_test"] = True + self.amp_eval = True + + # TODO(gaotingquan): to compatible with different versions of Paddle + paddle_version = paddle.__version__[:3] + # paddle version < 2.3.0 and not develop + if paddle_version not in ["2.3", "0.0"]: + if self.mode == "train": + self.model, self.optimizer = paddle.amp.decorate( + models=self.model, + optimizers=self.optimizer, + level=self.amp_level, + save_dtype='float32') + elif self.amp_eval: + if self.amp_level == "O2": + msg = "The PaddlePaddle that installed not support FP16 evaluation in AMP O2. Please use PaddlePaddle version >= 2.3.0. Use FP32 evaluation instead and please notice the Eval Dataset output_fp16 should be 'False'." + logger.warning(msg) + self.amp_eval = False + else: + self.model, self.optimizer = paddle.amp.decorate( + models=self.model, + level=self.amp_level, + save_dtype='float32') + # paddle version >= 2.3.0 or develop + else: + self.model = paddle.amp.decorate( + models=self.model, + level=self.amp_level, + save_dtype='float32') + + if self.mode == "train" and len(self.train_loss_func.parameters( + )) > 0: + self.train_loss_func = paddle.amp.decorate( + models=self.train_loss_func, + level=self.amp_level, + save_dtype='float32') + + # check the gpu num world_size = dist.get_world_size() self.config["Global"]["distributed"] = world_size != 1 - if world_size != 4 and self.mode == "train": - msg = f"The training strategy in config files provided by PaddleClas is based on 4 gpus. But the number of gpus is {world_size} in current training. Please modify the stategy (learning rate, batch size and so on) if use config files in PaddleClas to train." - logger.warning(msg) + if self.mode == "train": + std_gpu_num = 8 if isinstance( + self.config["Optimizer"], + dict) and self.config["Optimizer"]["name"] == "AdamW" else 4 + if world_size != std_gpu_num: + msg = f"The training strategy provided by PaddleClas is based on {std_gpu_num} gpus. But the number of gpu is {world_size} in current training. Please modify the stategy (learning rate, batch size and so on) if use this config to train." + logger.warning(msg) + + # for distributed if self.config["Global"]["distributed"]: dist.init_parallel_env() self.model = paddle.DataParallel(self.model) - + if self.mode == 'train' and len(self.train_loss_func.parameters( + )) > 0: + self.train_loss_func = paddle.DataParallel( + self.train_loss_func) # build postprocess for infer if self.mode == 'infer': self.preprocess_func = create_operators(self.config["Infer"][ @@ -275,14 +327,15 @@ class Engine(object): # global iter counter self.global_step = 0 - if self.config["Global"]["checkpoints"] is not None: - metric_info = init_model(self.config["Global"], self.model, - self.optimizer) + if self.config.Global.checkpoints is not None: + metric_info = init_model(self.config.Global, self.model, + self.optimizer, self.train_loss_func) if metric_info is not None: best_metric.update(metric_info) self.max_iter = len(self.train_dataloader) - 1 if platform.system( ) == "Windows" else len(self.train_dataloader) + for epoch_id in range(best_metric["epoch"] + 1, self.config["Global"]["epochs"] + 1): acc = 0.0 @@ -313,7 +366,8 @@ class Engine(object): best_metric, self.output_dir, model_name=self.config["Arch"]["name"], - prefix="best_model") + prefix="best_model", + loss=self.train_loss_func) logger.info("[Eval][Epoch {}][best metric: {}]".format( epoch_id, best_metric["metric"])) logger.scaler( @@ -332,7 +386,8 @@ class Engine(object): "epoch": epoch_id}, self.output_dir, model_name=self.config["Arch"]["name"], - prefix="epoch_{}".format(epoch_id)) + prefix="epoch_{}".format(epoch_id), + loss=self.train_loss_func) # save the latest model save_load.save_model( self.model, @@ -340,7 +395,8 @@ class Engine(object): "epoch": epoch_id}, self.output_dir, model_name=self.config["Arch"]["name"], - prefix="latest") + prefix="latest", + loss=self.train_loss_func) if self.vdl_writer is not None: self.vdl_writer.close() @@ -459,5 +515,7 @@ class ExportModel(TheseusLayer): if self.infer_output_key is not None: x = x[self.infer_output_key] if self.out_act is not None: + if isinstance(x, dict): + x = x["logits"] x = self.out_act(x) return x diff --git a/ppcls/engine/evaluation/classification.py b/ppcls/engine/evaluation/classification.py index d7b5c47620bcd03b9ef8ddd44deeea0621ca041d..60595e6a9014b4003ab8008b8144d92d628a2acd 100644 --- a/ppcls/engine/evaluation/classification.py +++ b/ppcls/engine/evaluation/classification.py @@ -53,79 +53,77 @@ def classification_eval(engine, epoch_id=0): ] time_info["reader_cost"].update(time.time() - tic) batch_size = batch[0].shape[0] - batch[0] = paddle.to_tensor(batch[0]).astype("float32") + batch[0] = paddle.to_tensor(batch[0]) if not engine.config["Global"].get("use_multilabel", False): batch[1] = batch[1].reshape([-1, 1]).astype("int64") # image input - if engine.amp: - amp_level = engine.config['AMP'].get("level", "O1").upper() + if engine.amp and engine.amp_eval: with paddle.amp.auto_cast( custom_black_list={ "flatten_contiguous_range", "greater_than" }, - level=amp_level): + level=engine.amp_level): out = engine.model(batch[0]) - # calc loss - if engine.eval_loss_func is not None: - loss_dict = engine.eval_loss_func(out, batch[1]) - for key in loss_dict: - if key not in output_info: - output_info[key] = AverageMeter(key, '7.5f') - output_info[key].update(loss_dict[key].numpy()[0], - batch_size) else: out = engine.model(batch[0]) - # calc loss - if engine.eval_loss_func is not None: - loss_dict = engine.eval_loss_func(out, batch[1]) - for key in loss_dict: - if key not in output_info: - output_info[key] = AverageMeter(key, '7.5f') - output_info[key].update(loss_dict[key].numpy()[0], - batch_size) # just for DistributedBatchSampler issue: repeat sampling current_samples = batch_size * paddle.distributed.get_world_size() accum_samples += current_samples - # calc metric - if engine.eval_metric_func is not None: - if paddle.distributed.get_world_size() > 1: - label_list = [] - paddle.distributed.all_gather(label_list, batch[1]) - labels = paddle.concat(label_list, 0) - - if isinstance(out, dict): - if "Student" in out: - out = out["Student"] - elif "logits" in out: - out = out["logits"] - else: - msg = "Error: Wrong key in out!" - raise Exception(msg) - if isinstance(out, list): - pred = [] - for x in out: - pred_list = [] - paddle.distributed.all_gather(pred_list, x) - pred_x = paddle.concat(pred_list, 0) - pred.append(pred_x) - else: + if isinstance(out, dict) and "Student" in out: + out = out["Student"] + if isinstance(out, dict) and "logits" in out: + out = out["logits"] + + # gather Tensor when distributed + if paddle.distributed.get_world_size() > 1: + label_list = [] + paddle.distributed.all_gather(label_list, batch[1]) + labels = paddle.concat(label_list, 0) + + if isinstance(out, list): + preds = [] + for x in out: pred_list = [] - paddle.distributed.all_gather(pred_list, out) - pred = paddle.concat(pred_list, 0) + paddle.distributed.all_gather(pred_list, x) + pred_x = paddle.concat(pred_list, 0) + preds.append(pred_x) + else: + pred_list = [] + paddle.distributed.all_gather(pred_list, out) + preds = paddle.concat(pred_list, 0) - if accum_samples > total_samples and not engine.use_dali: - pred = pred[:total_samples + current_samples - + if accum_samples > total_samples and not engine.use_dali: + preds = preds[:total_samples + current_samples - accum_samples] + labels = labels[:total_samples + current_samples - accum_samples] - labels = labels[:total_samples + current_samples - - accum_samples] - current_samples = total_samples + current_samples - accum_samples - metric_dict = engine.eval_metric_func(pred, labels) + current_samples = total_samples + current_samples - accum_samples + else: + labels = batch[1] + preds = out + + # calc loss + if engine.eval_loss_func is not None: + if engine.amp and engine.amp_eval: + with paddle.amp.auto_cast( + custom_black_list={ + "flatten_contiguous_range", "greater_than" + }, + level=engine.amp_level): + loss_dict = engine.eval_loss_func(preds, labels) else: - metric_dict = engine.eval_metric_func(out, batch[1]) + loss_dict = engine.eval_loss_func(preds, labels) + for key in loss_dict: + if key not in output_info: + output_info[key] = AverageMeter(key, '7.5f') + output_info[key].update(loss_dict[key].numpy()[0], + current_samples) + # calc metric + if engine.eval_metric_func is not None: + metric_dict = engine.eval_metric_func(preds, labels) for key in metric_dict: if metric_key is None: metric_key = key diff --git a/ppcls/engine/evaluation/retrieval.py b/ppcls/engine/evaluation/retrieval.py index 8471a42c7421648b75a56e426bf3e1ab9c14a5fd..05c5d0c35d0f6fdfcd0a8f1dc1a8a121026ede99 100644 --- a/ppcls/engine/evaluation/retrieval.py +++ b/ppcls/engine/evaluation/retrieval.py @@ -89,10 +89,8 @@ def retrieval_eval(engine, epoch_id=0): def cal_feature(engine, name='gallery'): - all_feas = None - all_image_id = None - all_unique_id = None has_unique_id = False + all_unique_id = None if name == 'gallery': dataloader = engine.gallery_dataloader @@ -103,6 +101,9 @@ def cal_feature(engine, name='gallery'): else: raise RuntimeError("Only support gallery or query dataset") + batch_feas_list = [] + img_id_list = [] + unique_id_list = [] max_iter = len(dataloader) - 1 if platform.system() == "Windows" else len( dataloader) for idx, batch in enumerate(dataloader): # load is very time-consuming @@ -125,7 +126,15 @@ def cal_feature(engine, name='gallery'): out = engine.model(batch[0], batch[1]) if "Student" in out: out = out["Student"] - batch_feas = out["features"] + + # get features + if engine.config["Global"].get("retrieval_feature_from", + "features") == "features": + # use neck's output as features + batch_feas = out["features"] + else: + # use backbone's output as features + batch_feas = out["backbone"] # do norm if engine.config["Global"].get("feature_normalize", True): @@ -140,32 +149,39 @@ def cal_feature(engine, name='gallery'): if engine.config["Global"].get("feature_binarize") == "sign": batch_feas = paddle.sign(batch_feas).astype("float32") - if all_feas is None: - all_feas = batch_feas + if paddle.distributed.get_world_size() > 1: + batch_feas_gather = [] + img_id_gather = [] + unique_id_gather = [] + paddle.distributed.all_gather(batch_feas_gather, batch_feas) + paddle.distributed.all_gather(img_id_gather, batch[1]) + batch_feas_list.append(paddle.concat(batch_feas_gather)) + img_id_list.append(paddle.concat(img_id_gather)) if has_unique_id: - all_unique_id = batch[2] - all_image_id = batch[1] + paddle.distributed.all_gather(unique_id_gather, batch[2]) + unique_id_list.append(paddle.concat(unique_id_gather)) else: - all_feas = paddle.concat([all_feas, batch_feas]) - all_image_id = paddle.concat([all_image_id, batch[1]]) + batch_feas_list.append(batch_feas) + img_id_list.append(batch[1]) if has_unique_id: - all_unique_id = paddle.concat([all_unique_id, batch[2]]) + unique_id_list.append(batch[2]) if engine.use_dali: dataloader.reset() - if paddle.distributed.get_world_size() > 1: - feat_list = [] - img_id_list = [] - unique_id_list = [] - paddle.distributed.all_gather(feat_list, all_feas) - paddle.distributed.all_gather(img_id_list, all_image_id) - all_feas = paddle.concat(feat_list, axis=0) - all_image_id = paddle.concat(img_id_list, axis=0) - if has_unique_id: - paddle.distributed.all_gather(unique_id_list, all_unique_id) - all_unique_id = paddle.concat(unique_id_list, axis=0) + all_feas = paddle.concat(batch_feas_list) + all_img_id = paddle.concat(img_id_list) + if has_unique_id: + all_unique_id = paddle.concat(unique_id_list) + + # just for DistributedBatchSampler issue: repeat sampling + total_samples = len( + dataloader.dataset) if not engine.use_dali else dataloader.size + all_feas = all_feas[:total_samples] + all_img_id = all_img_id[:total_samples] + if has_unique_id: + all_unique_id = all_unique_id[:total_samples] logger.info("Build {} done, all feat shape: {}, begin to eval..".format( name, all_feas.shape)) - return all_feas, all_image_id, all_unique_id + return all_feas, all_img_id, all_unique_id diff --git a/ppcls/engine/train/train.py b/ppcls/engine/train/train.py index 3b02bac8f305d9b944eb97a5a98ba3c39521191a..14db79e73e9e51d16d5784b7aa48a6afb12a7e0f 100644 --- a/ppcls/engine/train/train.py +++ b/ppcls/engine/train/train.py @@ -21,7 +21,6 @@ from ppcls.utils import profiler def train_epoch(engine, epoch_id, print_batch_step): tic = time.time() - v_current = [int(i) for i in paddle.__version__.split(".")] for iter_id, batch in enumerate(engine.train_dataloader): if iter_id >= engine.max_iter: break @@ -54,16 +53,25 @@ def train_epoch(engine, epoch_id, print_batch_step): out = forward(engine, batch) loss_dict = engine.train_loss_func(out, batch[1]) - # step opt and lr + # backward & step opt if engine.amp: scaled = engine.scaler.scale(loss_dict["loss"]) scaled.backward() - engine.scaler.minimize(engine.optimizer, scaled) + for i in range(len(engine.optimizer)): + engine.scaler.minimize(engine.optimizer[i], scaled) else: loss_dict["loss"].backward() - engine.optimizer.step() - engine.optimizer.clear_grad() - engine.lr_sch.step() + for i in range(len(engine.optimizer)): + engine.optimizer[i].step() + + # clear grad + for i in range(len(engine.optimizer)): + engine.optimizer[i].clear_grad() + + # step lr(by step) + for i in range(len(engine.lr_sch)): + if not getattr(engine.lr_sch[i], "by_epoch", False): + engine.lr_sch[i].step() # below code just for logging # update metric_for_logger @@ -75,6 +83,11 @@ def train_epoch(engine, epoch_id, print_batch_step): log_info(engine, batch_size, epoch_id, iter_id) tic = time.time() + # step lr(by epoch) + for i in range(len(engine.lr_sch)): + if getattr(engine.lr_sch[i], "by_epoch", False): + engine.lr_sch[i].step() + def forward(engine, batch): if not engine.is_rec: diff --git a/ppcls/engine/train/utils.py b/ppcls/engine/train/utils.py index 92eb35d7549c72a7ef3f4907d867e50d237c3f75..ca211ff932f19ca63804a5a1ff52def5eb89477f 100644 --- a/ppcls/engine/train/utils.py +++ b/ppcls/engine/train/utils.py @@ -38,7 +38,10 @@ def update_loss(trainer, loss_dict, batch_size): def log_info(trainer, batch_size, epoch_id, iter_id): - lr_msg = "lr: {:.5f}".format(trainer.lr_sch.get_lr()) + lr_msg = ", ".join([ + "lr({}): {:.8f}".format(lr.__class__.__name__, lr.get_lr()) + for i, lr in enumerate(trainer.lr_sch) + ]) metric_msg = ", ".join([ "{}: {:.5f}".format(key, trainer.output_info[key].avg) for key in trainer.output_info @@ -48,7 +51,7 @@ def log_info(trainer, batch_size, epoch_id, iter_id): for key in trainer.time_info ]) - ips_msg = "ips: {:.5f} images/sec".format( + ips_msg = "ips: {:.5f} samples/s".format( batch_size / trainer.time_info["batch_cost"].avg) eta_sec = ((trainer.config["Global"]["epochs"] - epoch_id + 1 ) * len(trainer.train_dataloader) - iter_id @@ -59,11 +62,12 @@ def log_info(trainer, batch_size, epoch_id, iter_id): len(trainer.train_dataloader), lr_msg, metric_msg, time_msg, ips_msg, eta_msg)) - logger.scaler( - name="lr", - value=trainer.lr_sch.get_lr(), - step=trainer.global_step, - writer=trainer.vdl_writer) + for i, lr in enumerate(trainer.lr_sch): + logger.scaler( + name="lr({})".format(lr.__class__.__name__), + value=lr.get_lr(), + step=trainer.global_step, + writer=trainer.vdl_writer) for key in trainer.output_info: logger.scaler( name="train_{}".format(key), diff --git a/ppcls/loss/__init__.py b/ppcls/loss/__init__.py index d15dab9da36c02c077b2b7f871d98f57582dbd85..c1f2f95df7afd0a266304ea2ccdf5572d1de9625 100644 --- a/ppcls/loss/__init__.py +++ b/ppcls/loss/__init__.py @@ -22,9 +22,14 @@ from .distillationloss import DistillationGTCELoss from .distillationloss import DistillationDMLLoss from .distillationloss import DistillationDistanceLoss from .distillationloss import DistillationRKDLoss +from .distillationloss import DistillationKLDivLoss +from .distillationloss import DistillationDKDLoss from .multilabelloss import MultiLabelLoss +from .afdloss import AFDLoss -from .deephashloss import DSHSDLoss, LCDSHLoss +from .deephashloss import DSHSDLoss +from .deephashloss import LCDSHLoss +from .deephashloss import DCHLoss class CombinedLoss(nn.Layer): @@ -43,6 +48,7 @@ class CombinedLoss(nn.Layer): param.keys()) self.loss_weight.append(param.pop("weight")) self.loss_func.append(eval(name)(**param)) + self.loss_func = nn.LayerList(self.loss_func) def __call__(self, input, batch): loss_dict = {} diff --git a/ppcls/loss/afdloss.py b/ppcls/loss/afdloss.py new file mode 100644 index 0000000000000000000000000000000000000000..3e67e30b98df61576e40449015cc67a13dd6da60 --- /dev/null +++ b/ppcls/loss/afdloss.py @@ -0,0 +1,132 @@ +#copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +#Unless required by applicable law or agreed to in writing, software +#distributed under the License is distributed on an "AS IS" BASIS, +#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +#See the License for the specific language governing permissions and +#limitations under the License. + +import paddle.nn as nn +import paddle.nn.functional as F +import paddle +import numpy as np +import matplotlib.pyplot as plt +import cv2 +import warnings +warnings.filterwarnings('ignore') + + +class LinearBNReLU(nn.Layer): + def __init__(self, nin, nout): + super().__init__() + self.linear = nn.Linear(nin, nout) + self.bn = nn.BatchNorm1D(nout) + self.relu = nn.ReLU() + + def forward(self, x, relu=True): + if relu: + return self.relu(self.bn(self.linear(x))) + return self.bn(self.linear(x)) + + +def unique_shape(s_shapes): + n_s = [] + unique_shapes = [] + n = -1 + for s_shape in s_shapes: + if s_shape not in unique_shapes: + unique_shapes.append(s_shape) + n += 1 + n_s.append(n) + return n_s, unique_shapes + + +class AFDLoss(nn.Layer): + """ + AFDLoss + https://www.aaai.org/AAAI21Papers/AAAI-9785.JiM.pdf + https://github.com/clovaai/attention-feature-distillation + """ + + def __init__(self, + model_name_pair=["Student", "Teacher"], + student_keys=["bilinear_key", "value"], + teacher_keys=["query", "value"], + s_shapes=[[64, 16, 160], [128, 8, 160], [256, 4, 160], + [512, 2, 160]], + t_shapes=[[640, 48], [320, 96], [160, 192]], + qk_dim=128, + name="loss_afd"): + super().__init__() + assert isinstance(model_name_pair, list) + self.model_name_pair = model_name_pair + self.student_keys = student_keys + self.teacher_keys = teacher_keys + self.s_shapes = [[1] + s_i for s_i in s_shapes] + self.t_shapes = [[1] + t_i for t_i in t_shapes] + self.qk_dim = qk_dim + self.n_t, self.unique_t_shapes = unique_shape(self.t_shapes) + self.attention = Attention(self.qk_dim, self.t_shapes, self.s_shapes, + self.n_t, self.unique_t_shapes) + self.name = name + + def forward(self, predicts, batch): + s_features_dict = predicts[self.model_name_pair[0]] + t_features_dict = predicts[self.model_name_pair[1]] + + g_s = [s_features_dict[key] for key in self.student_keys] + g_t = [t_features_dict[key] for key in self.teacher_keys] + + loss = self.attention(g_s, g_t) + sum_loss = sum(loss) + + loss_dict = dict() + loss_dict[self.name] = sum_loss + + return loss_dict + + +class Attention(nn.Layer): + def __init__(self, qk_dim, t_shapes, s_shapes, n_t, unique_t_shapes): + super().__init__() + self.qk_dim = qk_dim + self.n_t = n_t + # self.linear_trans_s = LinearTransformStudent(qk_dim, t_shapes, s_shapes, unique_t_shapes) + # self.linear_trans_t = LinearTransformTeacher(qk_dim, t_shapes) + + self.p_t = self.create_parameter( + shape=[len(t_shapes), qk_dim], + default_initializer=nn.initializer.XavierNormal()) + self.p_s = self.create_parameter( + shape=[len(s_shapes), qk_dim], + default_initializer=nn.initializer.XavierNormal()) + + def forward(self, g_s, g_t): + bilinear_key, h_hat_s_all = g_s + query, h_t_all = g_t + + p_logit = paddle.matmul(self.p_t, self.p_s.t()) + + logit = paddle.add( + paddle.einsum('bstq,btq->bts', bilinear_key, query), + p_logit) / np.sqrt(self.qk_dim) + atts = F.softmax(logit, axis=2) # b x t x s + + loss = [] + + for i, (n, h_t) in enumerate(zip(self.n_t, h_t_all)): + h_hat_s = h_hat_s_all[n] + diff = self.cal_diff(h_hat_s, h_t, atts[:, i]) + loss.append(diff) + return loss + + def cal_diff(self, v_s, v_t, att): + diff = (v_s - v_t.unsqueeze(1)).pow(2).mean(2) + diff = paddle.multiply(diff, att).sum(1).mean() + return diff diff --git a/ppcls/loss/centerloss.py b/ppcls/loss/centerloss.py index d85b3f2a90c781c2fdabf57ca852140c5a1090ba..23a86ee8875c1863beae749ea873f4cb662510d0 100644 --- a/ppcls/loss/centerloss.py +++ b/ppcls/loss/centerloss.py @@ -1,54 +1,80 @@ +# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + from __future__ import absolute_import from __future__ import division from __future__ import print_function + +from typing import Dict + import paddle import paddle.nn as nn -import paddle.nn.functional as F class CenterLoss(nn.Layer): - def __init__(self, num_classes=5013, feat_dim=2048): + """Center loss + paper : [A Discriminative Feature Learning Approach for Deep Face Recognition](https://link.springer.com/content/pdf/10.1007%2F978-3-319-46478-7_31.pdf) + code reference: https://github.com/michuanhaohao/reid-strong-baseline/blob/master/layers/center_loss.py#L7 + Args: + num_classes (int): number of classes. + feat_dim (int): number of feature dimensions. + feature_from (str): feature from "backbone" or "features" + """ + + def __init__(self, + num_classes: int, + feat_dim: int, + feature_from: str="features"): super(CenterLoss, self).__init__() self.num_classes = num_classes self.feat_dim = feat_dim - self.centers = paddle.randn( - shape=[self.num_classes, self.feat_dim]).astype( - "float64") #random center + self.feature_from = feature_from + random_init_centers = paddle.randn( + shape=[self.num_classes, self.feat_dim]) + self.centers = self.create_parameter( + shape=(self.num_classes, self.feat_dim), + default_initializer=nn.initializer.Assign(random_init_centers)) + self.add_parameter("centers", self.centers) - def __call__(self, input, target): - """ - inputs: network output: {"features: xxx", "logits": xxxx} - target: image label + def __call__(self, input: Dict[str, paddle.Tensor], + target: paddle.Tensor) -> Dict[str, paddle.Tensor]: + """compute center loss. + + Args: + input (Dict[str, paddle.Tensor]): {'features': (batch_size, feature_dim), ...}. + target (paddle.Tensor): ground truth label with shape (batch_size, ). + + Returns: + Dict[str, paddle.Tensor]: {'CenterLoss': loss}. """ - feats = input["features"] + feats = input[self.feature_from] labels = target + + # squeeze labels to shape (batch_size, ) + if labels.ndim >= 2 and labels.shape[-1] == 1: + labels = paddle.squeeze(labels, axis=[-1]) + batch_size = feats.shape[0] + distmat = paddle.pow(feats, 2).sum(axis=1, keepdim=True).expand([batch_size, self.num_classes]) + \ + paddle.pow(self.centers, 2).sum(axis=1, keepdim=True).expand([self.num_classes, batch_size]).t() + distmat = distmat.addmm(x=feats, y=self.centers.t(), beta=1, alpha=-2) - #calc feat * feat - dist1 = paddle.sum(paddle.square(feats), axis=1, keepdim=True) - dist1 = paddle.expand(dist1, [batch_size, self.num_classes]) - - #dist2 of centers - dist2 = paddle.sum(paddle.square(self.centers), axis=1, - keepdim=True) #num_classes - dist2 = paddle.expand(dist2, - [self.num_classes, batch_size]).astype("float64") - dist2 = paddle.transpose(dist2, [1, 0]) - - #first x * x + y * y - distmat = paddle.add(dist1, dist2) - tmp = paddle.matmul(feats, paddle.transpose(self.centers, [1, 0])) - distmat = distmat - 2.0 * tmp - - #generate the mask - classes = paddle.arange(self.num_classes).astype("int64") - labels = paddle.expand( - paddle.unsqueeze(labels, 1), (batch_size, self.num_classes)) - mask = paddle.equal( - paddle.expand(classes, [batch_size, self.num_classes]), - labels).astype("float64") #get mask - - dist = paddle.multiply(distmat, mask) - loss = paddle.sum(paddle.clip(dist, min=1e-12, max=1e+12)) / batch_size + classes = paddle.arange(self.num_classes).astype(labels.dtype) + labels = labels.unsqueeze(1).expand([batch_size, self.num_classes]) + mask = labels.equal(classes.expand([batch_size, self.num_classes])) + dist = distmat * mask.astype(feats.dtype) + loss = dist.clip(min=1e-12, max=1e+12).sum() / batch_size + # return loss return {'CenterLoss': loss} diff --git a/ppcls/loss/deephashloss.py b/ppcls/loss/deephashloss.py index c9a58dc78db48a0eef83a5aec2efb4b99d44ea91..7dda519a871cc0460160ea4411c286ac5c25d6ad 100644 --- a/ppcls/loss/deephashloss.py +++ b/ppcls/loss/deephashloss.py @@ -15,78 +15,135 @@ import paddle import paddle.nn as nn + class DSHSDLoss(nn.Layer): """ # DSHSD(IEEE ACCESS 2019) # paper [Deep Supervised Hashing Based on Stable Distribution](https://ieeexplore.ieee.org/document/8648432/) - # [DSHSD] epoch:70, bit:48, dataset:cifar10-1, MAP:0.809, Best MAP: 0.809 - # [DSHSD] epoch:250, bit:48, dataset:nuswide_21, MAP:0.809, Best MAP: 0.815 - # [DSHSD] epoch:135, bit:48, dataset:imagenet, MAP:0.647, Best MAP: 0.647 + # code reference: https://github.com/swuxyj/DeepHash-pytorch/blob/master/DSHSD.py """ + def __init__(self, alpha, multi_label=False): super(DSHSDLoss, self).__init__() self.alpha = alpha self.multi_label = multi_label def forward(self, input, label): - feature = input["features"] + features = input["features"] logits = input["logits"] - dist = paddle.sum(paddle.square( - (paddle.unsqueeze(feature, 1) - paddle.unsqueeze(feature, 0))), - axis=2) + features_temp1 = paddle.unsqueeze(features, 1) + features_temp2 = paddle.unsqueeze(features, 0) + dist = features_temp1 - features_temp2 + dist = paddle.square(dist) + dist = paddle.sum(dist, axis=2) - # label to ont-hot - label = paddle.flatten(label) n_class = logits.shape[1] - label = paddle.nn.functional.one_hot(label, n_class).astype("float32") + labels = paddle.nn.functional.one_hot(label, n_class) + labels = labels.squeeze().astype("float32") - s = (paddle.matmul( - label, label, transpose_y=True) == 0).astype("float32") - margin = 2 * feature.shape[1] + s = paddle.matmul(labels, labels, transpose_y=True) + s = (s == 0).astype("float32") + margin = 2 * features.shape[1] Ld = (1 - s) / 2 * dist + s / 2 * (margin - dist).clip(min=0) Ld = Ld.mean() if self.multi_label: - # multiple labels classification loss - Lc = (logits - label * logits + ( - (1 + (-logits).exp()).log())).sum(axis=1).mean() + Lc_temp = (1 + (-logits).exp()).log() + Lc = (logits - labels * logits + Lc_temp).sum(axis=1) else: - # single labels classification loss - Lc = (-paddle.nn.functional.softmax(logits).log() * label).sum( - axis=1).mean() + probs = paddle.nn.functional.softmax(logits) + Lc = (-probs.log() * labels).sum(axis=1) + Lc = Lc.mean() - return {"dshsdloss": Lc + Ld * self.alpha} + loss = Lc + Ld * self.alpha + return {"dshsdloss": loss} class LCDSHLoss(nn.Layer): """ # paper [Locality-Constrained Deep Supervised Hashing for Image Retrieval](https://www.ijcai.org/Proceedings/2017/0499.pdf) - # [LCDSH] epoch:145, bit:48, dataset:cifar10-1, MAP:0.798, Best MAP: 0.798 - # [LCDSH] epoch:183, bit:48, dataset:nuswide_21, MAP:0.833, Best MAP: 0.834 + # code reference: https://github.com/swuxyj/DeepHash-pytorch/blob/master/LCDSH.py """ + def __init__(self, n_class, _lambda): super(LCDSHLoss, self).__init__() self._lambda = _lambda self.n_class = n_class def forward(self, input, label): - feature = input["features"] + features = input["features"] + labels = paddle.nn.functional.one_hot(label, self.n_class) + labels = labels.squeeze().astype("float32") - # label to ont-hot - label = paddle.flatten(label) - label = paddle.nn.functional.one_hot(label, self.n_class).astype("float32") - - s = 2 * (paddle.matmul(label, label, transpose_y=True) > 0).astype("float32") - 1 - inner_product = paddle.matmul(feature, feature, transpose_y=True) * 0.5 + s = paddle.matmul(labels, labels, transpose_y=True) + s = 2 * (s > 0).astype("float32") - 1 + inner_product = paddle.matmul(features, features, transpose_y=True) + inner_product = inner_product * 0.5 inner_product = inner_product.clip(min=-50, max=50) - L1 = paddle.log(1 + paddle.exp(-s * inner_product)).mean() + L1 = paddle.log(1 + paddle.exp(-s * inner_product)) + L1 = L1.mean() - b = feature.sign() - inner_product_ = paddle.matmul(b, b, transpose_y=True) * 0.5 + binary_features = features.sign() + + inner_product_ = paddle.matmul( + binary_features, binary_features, transpose_y=True) + inner_product_ = inner_product_ * 0.5 sigmoid = paddle.nn.Sigmoid() - L2 = (sigmoid(inner_product) - sigmoid(inner_product_)).pow(2).mean() + L2 = (sigmoid(inner_product) - sigmoid(inner_product_)).pow(2) + L2 = L2.mean() + + loss = L1 + self._lambda * L2 + return {"lcdshloss": loss} + + +class DCHLoss(paddle.nn.Layer): + """ + # paper [Deep Cauchy Hashing for Hamming Space Retrieval] + URL:(http://ise.thss.tsinghua.edu.cn/~mlong/doc/deep-cauchy-hashing-cvpr18.pdf) + # code reference: https://github.com/swuxyj/DeepHash-pytorch/blob/master/DCH.py + """ + + def __init__(self, gamma, _lambda, n_class): + super(DCHLoss, self).__init__() + self.gamma = gamma + self._lambda = _lambda + self.n_class = n_class + + def distance(self, feature_i, feature_j): + assert feature_i.shape[1] == feature_j.shape[ + 1], "feature len of feature_i and feature_j is different, please check whether the featurs are right" + K = feature_i.shape[1] + inner_product = paddle.matmul(feature_i, feature_j, transpose_y=True) + + len_i = feature_i.pow(2).sum(axis=1, keepdim=True).pow(0.5) + len_j = feature_j.pow(2).sum(axis=1, keepdim=True).pow(0.5) + norm = paddle.matmul(len_i, len_j, transpose_y=True) + cos = inner_product / norm.clip(min=0.0001) + dist = (1 - cos.clip(max=0.99)) * K / 2 + return dist + + def forward(self, input, label): + features = input["features"] + labels = paddle.nn.functional.one_hot(label, self.n_class) + labels = labels.squeeze().astype("float32") + + s = paddle.matmul(labels, labels, transpose_y=True).astype("float32") + if (1 - s).sum() != 0 and s.sum() != 0: + positive_w = s * s.numel() / s.sum() + negative_w = (1 - s) * s.numel() / (1 - s).sum() + w = positive_w + negative_w + else: + w = 1 + + dist_matric = self.distance(features, features) + cauchy_loss = w * (s * paddle.log(dist_matric / self.gamma) + + paddle.log(1 + self.gamma / dist_matric)) - return {"lcdshloss": L1 + self._lambda * L2} + all_one = paddle.ones_like(features, dtype="float32") + dist_to_one = self.distance(features.abs(), all_one) + quantization_loss = paddle.log(1 + dist_to_one / self.gamma) + loss = cauchy_loss.mean() + self._lambda * quantization_loss.mean() + return {"dchloss": loss} diff --git a/ppcls/loss/distillationloss.py b/ppcls/loss/distillationloss.py index 0340234b9d2374adc97aa4be8b8c2e50ce297e6f..c60a540db84edae1374e5370309256f1c98cd40a 100644 --- a/ppcls/loss/distillationloss.py +++ b/ppcls/loss/distillationloss.py @@ -14,11 +14,14 @@ import paddle import paddle.nn as nn +import paddle.nn.functional as F from .celoss import CELoss from .dmlloss import DMLLoss from .distanceloss import DistanceLoss from .rkdloss import RKdAngle, RkdDistance +from .kldivloss import KLDivLoss +from .dkdloss import DKDLoss class DistillationCELoss(CELoss): @@ -172,3 +175,63 @@ class DistillationRKDLoss(nn.Layer): student_out, teacher_out) return loss_dict + + +class DistillationKLDivLoss(KLDivLoss): + """ + DistillationKLDivLoss + """ + + def __init__(self, + model_name_pairs=[], + temperature=4, + key=None, + name="loss_kl"): + super().__init__(temperature=temperature) + assert isinstance(model_name_pairs, list) + self.key = key + self.model_name_pairs = model_name_pairs + self.name = name + + def forward(self, predicts, batch): + loss_dict = dict() + for idx, pair in enumerate(self.model_name_pairs): + out1 = predicts[pair[0]] + out2 = predicts[pair[1]] + if self.key is not None: + out1 = out1[self.key] + out2 = out2[self.key] + loss = super().forward(out1, out2) + for key in loss: + loss_dict["{}_{}_{}".format(key, pair[0], pair[1])] = loss[key] + return loss_dict + + +class DistillationDKDLoss(DKDLoss): + """ + DistillationDKDLoss + """ + + def __init__(self, + model_name_pairs=[], + key=None, + temperature=1.0, + alpha=1.0, + beta=1.0, + name="loss_dkd"): + super().__init__(temperature=temperature, alpha=alpha, beta=beta) + self.key = key + self.model_name_pairs = model_name_pairs + self.name = name + + def forward(self, predicts, batch): + loss_dict = dict() + for idx, pair in enumerate(self.model_name_pairs): + out1 = predicts[pair[0]] + out2 = predicts[pair[1]] + if self.key is not None: + out1 = out1[self.key] + out2 = out2[self.key] + loss = super().forward(out1, out2, batch) + loss_dict[f"{self.name}_{pair[0]}_{pair[1]}"] = loss + return loss_dict diff --git a/ppcls/loss/dkdloss.py b/ppcls/loss/dkdloss.py new file mode 100644 index 0000000000000000000000000000000000000000..9ce2c56d9334697d784ebc0371d4d59120790154 --- /dev/null +++ b/ppcls/loss/dkdloss.py @@ -0,0 +1,61 @@ +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class DKDLoss(nn.Layer): + """ + DKDLoss + Reference: https://arxiv.org/abs/2203.08679 + Code was heavily based on https://github.com/megvii-research/mdistiller + """ + + def __init__(self, temperature=1.0, alpha=1.0, beta=1.0): + super().__init__() + self.temperature = temperature + self.alpha = alpha + self.beta = beta + + def forward(self, logits_student, logits_teacher, target): + gt_mask = _get_gt_mask(logits_student, target) + other_mask = 1 - gt_mask + pred_student = F.softmax(logits_student / self.temperature, axis=1) + pred_teacher = F.softmax(logits_teacher / self.temperature, axis=1) + pred_student = cat_mask(pred_student, gt_mask, other_mask) + pred_teacher = cat_mask(pred_teacher, gt_mask, other_mask) + log_pred_student = paddle.log(pred_student) + tckd_loss = (F.kl_div( + log_pred_student, pred_teacher, + reduction='sum') * (self.temperature**2) / target.shape[0]) + pred_teacher_part2 = F.softmax( + logits_teacher / self.temperature - 1000.0 * gt_mask, axis=1) + log_pred_student_part2 = F.log_softmax( + logits_student / self.temperature - 1000.0 * gt_mask, axis=1) + nckd_loss = (F.kl_div( + log_pred_student_part2, pred_teacher_part2, + reduction='sum') * (self.temperature**2) / target.shape[0]) + return self.alpha * tckd_loss + self.beta * nckd_loss + + +def _get_gt_mask(logits, target): + target = target.reshape([-1]).unsqueeze(1) + updates = paddle.ones_like(target) + mask = scatter( + paddle.zeros_like(logits), target, updates.astype('float32')) + return mask + + +def cat_mask(t, mask1, mask2): + t1 = (t * mask1).sum(axis=1, keepdim=True) + t2 = (t * mask2).sum(axis=1, keepdim=True) + rt = paddle.concat([t1, t2], axis=1) + return rt + + +def scatter(x, index, updates): + i, j = index.shape + grid_x, grid_y = paddle.meshgrid(paddle.arange(i), paddle.arange(j)) + index = paddle.stack([grid_x.flatten(), index.flatten()], axis=1) + updates_index = paddle.stack([grid_x.flatten(), grid_y.flatten()], axis=1) + updates = paddle.gather_nd(updates, index=updates_index) + return paddle.scatter_nd_add(x, index, updates) diff --git a/ppcls/loss/emlloss.py b/ppcls/loss/emlloss.py index 973570389ac08e11b47449fbefbaa9e5e8e33c83..38b707fe1a4eb9ed6d130a3eb9bc4f8762d4c189 100644 --- a/ppcls/loss/emlloss.py +++ b/ppcls/loss/emlloss.py @@ -23,6 +23,11 @@ from .comfunc import rerange_index class EmlLoss(paddle.nn.Layer): + """Ensemble Metric Learning Loss + paper: [Large Scale Strongly Supervised Ensemble Metric Learning, with Applications to Face Verification and Retrieval](https://arxiv.org/pdf/1212.6094.pdf) + code reference: https://github.com/PaddlePaddle/models/blob/develop/PaddleCV/metric_learning/losses/emlloss.py + """ + def __init__(self, batch_size=40, samples_each_class=2): super(EmlLoss, self).__init__() assert (batch_size % samples_each_class == 0) diff --git a/ppcls/loss/googlenetloss.py b/ppcls/loss/googlenetloss.py index c580aa61701be6f5b6be43ce52a31be363b40d95..491311831acf90e11474f0a82713ef096221eb7f 100644 --- a/ppcls/loss/googlenetloss.py +++ b/ppcls/loss/googlenetloss.py @@ -18,11 +18,13 @@ import paddle.nn.functional as F class GoogLeNetLoss(nn.Layer): """ Cross entropy loss used after googlenet + reference paper: [https://arxiv.org/pdf/1409.4842v1.pdf](Going Deeper with Convolutions) """ + def __init__(self, epsilon=None): super().__init__() - assert (epsilon is None or epsilon <= 0 or epsilon >= 1), "googlenet is not support label_smooth" - + assert (epsilon is None or epsilon <= 0 or + epsilon >= 1), "googlenet is not support label_smooth" def forward(self, inputs, label): input0, input1, input2 = inputs diff --git a/ppcls/loss/kldivloss.py b/ppcls/loss/kldivloss.py new file mode 100644 index 0000000000000000000000000000000000000000..da6ab02fbb06874e0b54a6985f3c82f95db68979 --- /dev/null +++ b/ppcls/loss/kldivloss.py @@ -0,0 +1,33 @@ +# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import paddle +import paddle.nn as nn +import paddle.nn.functional as F + + +class KLDivLoss(nn.Layer): + """ + Distilling the Knowledge in a Neural Network + """ + + def __init__(self, temperature=4): + super(KLDivLoss, self).__init__() + self.T = temperature + + def forward(self, y_s, y_t): + p_s = F.log_softmax(y_s / self.T, axis=1) + p_t = F.softmax(y_t / self.T, axis=1) + loss = F.kl_div(p_s, p_t, reduction='sum') * (self.T**2) / y_s.shape[0] + return {"loss_kldiv": loss} diff --git a/ppcls/loss/msmloss.py b/ppcls/loss/msmloss.py index 3aa0dd8bfb0cdc6f558ff9891f0e0000ef183fae..adf03ef8e03c942fd1f2635704b9929e439dc3f5 100644 --- a/ppcls/loss/msmloss.py +++ b/ppcls/loss/msmloss.py @@ -21,10 +21,12 @@ from .comfunc import rerange_index class MSMLoss(paddle.nn.Layer): """ - MSMLoss Loss, based on triplet loss. USE P * K samples. + paper : [Margin Sample Mining Loss: A Deep Learning Based Method for Person Re-identification](https://arxiv.org/pdf/1710.00478.pdf) + code reference: https://github.com/michuanhaohao/keras_reid/blob/master/reid_tripletcls.py + Margin Sample Mining Loss, based on triplet loss. USE P * K samples. the batch size is fixed. Batch_size = P * K; but the K may vary between batches. same label gather together - + supported_metrics = [ 'euclidean', 'sqeuclidean', @@ -41,7 +43,7 @@ class MSMLoss(paddle.nn.Layer): self.rerange_index = rerange_index(batch_size, samples_each_class) def forward(self, input, target=None): - #normalization + #normalization features = input["features"] features = self._nomalize(features) samples_each_class = self.samples_each_class @@ -53,7 +55,7 @@ class MSMLoss(paddle.nn.Layer): features, axis=0) similary_matrix = paddle.sum(paddle.square(diffs), axis=-1) - #rerange + #rerange tmp = paddle.reshape(similary_matrix, shape=[-1, 1]) tmp = paddle.gather(tmp, index=rerange_index) similary_matrix = paddle.reshape(tmp, shape=[-1, self.batch_size]) diff --git a/ppcls/loss/npairsloss.py b/ppcls/loss/npairsloss.py index d4b359e88119a735442858cb8dbe3fa255add09a..131c799a48abb9507cfe7ae16dd2aa34bf8c8f25 100644 --- a/ppcls/loss/npairsloss.py +++ b/ppcls/loss/npairsloss.py @@ -5,6 +5,11 @@ import paddle class NpairsLoss(paddle.nn.Layer): + """Npair_loss_ + paper [Improved deep metric learning with multi-class N-pair loss objective](https://dl.acm.org/doi/10.5555/3157096.3157304) + code reference: https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/contrib/losses/metric_learning/npairs_loss + """ + def __init__(self, reg_lambda=0.01): super(NpairsLoss, self).__init__() self.reg_lambda = reg_lambda diff --git a/ppcls/loss/pairwisecosface.py b/ppcls/loss/pairwisecosface.py index beb806863bd171635452fd67993cc56404efe0b6..7f146dea5454b90e08a4790b2d95d4624c76bb0d 100644 --- a/ppcls/loss/pairwisecosface.py +++ b/ppcls/loss/pairwisecosface.py @@ -23,6 +23,11 @@ import paddle.nn.functional as F class PairwiseCosface(nn.Layer): + """ + paper: Circle Loss: A Unified Perspective of Pair Similarity Optimization + code reference: https://github.com/leoluopy/circle-loss-demonstration/blob/main/circle_loss.py + """ + def __init__(self, margin, gamma): super(PairwiseCosface, self).__init__() self.margin = margin @@ -36,8 +41,10 @@ class PairwiseCosface(nn.Layer): dist_mat = paddle.matmul(embedding, embedding, transpose_y=True) N = dist_mat.shape[0] - is_pos = targets.reshape([N,1]).expand([N,N]).equal(paddle.t(targets.reshape([N,1]).expand([N,N]))).astype('float') - is_neg = targets.reshape([N,1]).expand([N,N]).not_equal(paddle.t(targets.reshape([N,1]).expand([N,N]))).astype('float') + is_pos = targets.reshape([N, 1]).expand([N, N]).equal( + paddle.t(targets.reshape([N, 1]).expand([N, N]))).astype('float') + is_neg = targets.reshape([N, 1]).expand([N, N]).not_equal( + paddle.t(targets.reshape([N, 1]).expand([N, N]))).astype('float') # Mask scores related to itself is_pos = is_pos - paddle.eye(N, N) @@ -46,10 +53,12 @@ class PairwiseCosface(nn.Layer): s_n = dist_mat * is_neg logit_p = -self.gamma * s_p + (-99999999.) * (1 - is_pos) - logit_n = self.gamma * (s_n + self.margin) + (-99999999.) * (1 - is_neg) - - loss = F.softplus(paddle.logsumexp(logit_p, axis=1) + paddle.logsumexp(logit_n, axis=1)).mean() - - return {"PairwiseCosface": loss} + logit_n = self.gamma * (s_n + self.margin) + (-99999999.) * (1 - is_neg + ) + loss = F.softplus( + paddle.logsumexp( + logit_p, axis=1) + paddle.logsumexp( + logit_n, axis=1)).mean() + return {"PairwiseCosface": loss} diff --git a/ppcls/loss/rkdloss.py b/ppcls/loss/rkdloss.py index e6ffea273431ec7105d0cdedd0225c40648d2660..aa6ae232438ed6d5a915ea982092f3711d2901c3 100644 --- a/ppcls/loss/rkdloss.py +++ b/ppcls/loss/rkdloss.py @@ -29,6 +29,7 @@ def pdist(e, squared=False, eps=1e-12): class RKdAngle(nn.Layer): + # paper : [Relational Knowledge Distillation](https://arxiv.org/abs/1904.05068?context=cs.LG) # reference: https://github.com/lenscloth/RKD/blob/master/metric/loss.py def __init__(self, target_size=None): super().__init__() @@ -64,6 +65,7 @@ class RKdAngle(nn.Layer): class RkdDistance(nn.Layer): + # paper : [Relational Knowledge Distillation](https://arxiv.org/abs/1904.05068?context=cs.LG) # reference: https://github.com/lenscloth/RKD/blob/master/metric/loss.py def __init__(self, eps=1e-12, target_size=1): super().__init__() diff --git a/ppcls/loss/supconloss.py b/ppcls/loss/supconloss.py index 3dd33bc19e97ddb29966f55c4789b7a4ae81422b..753ceaf415b28793ebd1758c9cde84316d04e70b 100644 --- a/ppcls/loss/supconloss.py +++ b/ppcls/loss/supconloss.py @@ -4,6 +4,7 @@ from paddle import nn class SupConLoss(nn.Layer): """Supervised Contrastive Learning: https://arxiv.org/pdf/2004.11362.pdf. + code reference: https://github.com/HobbitLong/SupContrast/blob/master/losses.py It also supports the unsupervised contrastive loss in SimCLR""" def __init__(self, diff --git a/ppcls/loss/trihardloss.py b/ppcls/loss/trihardloss.py index 132c604d51920786e89c331b3223884910e50fa8..96cb42cb46b018abc77286eb96ef5e44f20f67b0 100644 --- a/ppcls/loss/trihardloss.py +++ b/ppcls/loss/trihardloss.py @@ -22,10 +22,12 @@ from .comfunc import rerange_index class TriHardLoss(paddle.nn.Layer): """ + paper: In Defense of the Triplet Loss for Person Re-Identification + code reference: https://github.com/VisualComputingInstitute/triplet-reid/blob/master/loss.py TriHard Loss, based on triplet loss. USE P * K samples. the batch size is fixed. Batch_size = P * K; but the K may vary between batches. same label gather together - + supported_metrics = [ 'euclidean', 'sqeuclidean', @@ -45,7 +47,7 @@ class TriHardLoss(paddle.nn.Layer): features = input["features"] assert (self.batch_size == features.shape[0]) - #normalization + #normalization features = self._nomalize(features) samples_each_class = self.samples_each_class rerange_index = paddle.to_tensor(self.rerange_index) @@ -56,7 +58,7 @@ class TriHardLoss(paddle.nn.Layer): features, axis=0) similary_matrix = paddle.sum(paddle.square(diffs), axis=-1) - #rerange + #rerange tmp = paddle.reshape(similary_matrix, shape=[-1, 1]) tmp = paddle.gather(tmp, index=rerange_index) similary_matrix = paddle.reshape(tmp, shape=[-1, self.batch_size]) diff --git a/ppcls/loss/triplet.py b/ppcls/loss/triplet.py index d1c7eec9e6031aa7e51a1a3575094e7d1a4f90df..0da7cc5dffb8f54807fa3d4da12b002755e54452 100644 --- a/ppcls/loss/triplet.py +++ b/ppcls/loss/triplet.py @@ -1,3 +1,17 @@ +# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + from __future__ import absolute_import from __future__ import division from __future__ import print_function @@ -8,13 +22,19 @@ import paddle.nn as nn class TripletLossV2(nn.Layer): """Triplet loss with hard positive/negative mining. + paper : [Facenet: A unified embedding for face recognition and clustering](https://arxiv.org/pdf/1503.03832.pdf) + code reference: https://github.com/okzhili/Cartoon-face-recognition/blob/master/loss/triplet_loss.py Args: margin (float): margin for triplet. """ - def __init__(self, margin=0.5, normalize_feature=True): + def __init__(self, + margin=0.5, + normalize_feature=True, + feature_from="features"): super(TripletLossV2, self).__init__() self.margin = margin + self.feature_from = feature_from self.ranking_loss = paddle.nn.loss.MarginRankingLoss(margin=margin) self.normalize_feature = normalize_feature @@ -24,7 +44,7 @@ class TripletLossV2(nn.Layer): inputs: feature matrix with shape (batch_size, feat_dim) target: ground truth labels with shape (num_classes) """ - inputs = input["features"] + inputs = input[self.feature_from] if self.normalize_feature: inputs = 1. * inputs / (paddle.expand_as( diff --git a/ppcls/metric/metrics.py b/ppcls/metric/metrics.py index 7c6407e7a4c74fa7d4330d72c6be52f6a843cdf0..03e742082b57439227746d21695379b498e7f1d8 100644 --- a/ppcls/metric/metrics.py +++ b/ppcls/metric/metrics.py @@ -302,8 +302,5 @@ class AccuracyScore(MutiLabelMetric): fps = mcm[:, 0, 1] accuracy = (sum(tps) + sum(tns)) / ( sum(tps) + sum(tns) + sum(fns) + sum(fps)) - precision = sum(tps) / (sum(tps) + sum(fps)) - recall = sum(tps) / (sum(tps) + sum(fns)) - F1 = 2 * (accuracy * recall) / (accuracy + recall) metric_dict["AccuracyScore"] = paddle.to_tensor(accuracy) return metric_dict diff --git a/ppcls/optimizer/__init__.py b/ppcls/optimizer/__init__.py index 61db39f89a3775eec70fe3cfa63d7e820e3e60e6..bdee9f9b6c4b605a85b635f6a12de5eda6165c90 100644 --- a/ppcls/optimizer/__init__.py +++ b/ppcls/optimizer/__init__.py @@ -18,6 +18,7 @@ from __future__ import print_function import copy import paddle +from typing import Dict, List from ppcls.utils import logger @@ -43,30 +44,87 @@ def build_lr_scheduler(lr_config, epochs, step_each_epoch): # model_list is None in static graph def build_optimizer(config, epochs, step_each_epoch, model_list=None): - config = copy.deepcopy(config) - # step1 build lr - lr = build_lr_scheduler(config.pop('lr'), epochs, step_each_epoch) - logger.debug("build lr ({}) success..".format(lr)) - # step2 build regularization - if 'regularizer' in config and config['regularizer'] is not None: - if 'weight_decay' in config: - logger.warning( - "ConfigError: Only one of regularizer and weight_decay can be set in Optimizer Config. \"weight_decay\" has been ignored." - ) - reg_config = config.pop('regularizer') - reg_name = reg_config.pop('name') + 'Decay' - reg = getattr(paddle.regularizer, reg_name)(**reg_config) - config["weight_decay"] = reg - logger.debug("build regularizer ({}) success..".format(reg)) - # step3 build optimizer - optim_name = config.pop('name') - if 'clip_norm' in config: - clip_norm = config.pop('clip_norm') - grad_clip = paddle.nn.ClipGradByNorm(clip_norm=clip_norm) - else: - grad_clip = None - optim = getattr(optimizer, optim_name)(learning_rate=lr, - grad_clip=grad_clip, - **config)(model_list=model_list) - logger.debug("build optimizer ({}) success..".format(optim)) - return optim, lr + optim_config = copy.deepcopy(config) + if isinstance(optim_config, dict): + # convert {'name': xxx, **optim_cfg} to [{name: {scope: xxx, **optim_cfg}}] + optim_name = optim_config.pop("name") + optim_config: List[Dict[str, Dict]] = [{ + optim_name: { + 'scope': "all", + ** + optim_config + } + }] + optim_list = [] + lr_list = [] + """NOTE: + Currently only support optim objets below. + 1. single optimizer config. + 2. next level uner Arch, such as Arch.backbone, Arch.neck, Arch.head. + 3. loss which has parameters, such as CenterLoss. + """ + for optim_item in optim_config: + # optim_cfg = {optim_name: {scope: xxx, **optim_cfg}} + # step1 build lr + optim_name = list(optim_item.keys())[0] # get optim_name + optim_scope = optim_item[optim_name].pop('scope') # get optim_scope + optim_cfg = optim_item[optim_name] # get optim_cfg + + lr = build_lr_scheduler(optim_cfg.pop('lr'), epochs, step_each_epoch) + logger.debug("build lr ({}) for scope ({}) success..".format( + lr, optim_scope)) + # step2 build regularization + if 'regularizer' in optim_cfg and optim_cfg['regularizer'] is not None: + if 'weight_decay' in optim_cfg: + logger.warning( + "ConfigError: Only one of regularizer and weight_decay can be set in Optimizer Config. \"weight_decay\" has been ignored." + ) + reg_config = optim_cfg.pop('regularizer') + reg_name = reg_config.pop('name') + 'Decay' + reg = getattr(paddle.regularizer, reg_name)(**reg_config) + optim_cfg["weight_decay"] = reg + logger.debug("build regularizer ({}) for scope ({}) success..". + format(reg, optim_scope)) + # step3 build optimizer + if 'clip_norm' in optim_cfg: + clip_norm = optim_cfg.pop('clip_norm') + grad_clip = paddle.nn.ClipGradByNorm(clip_norm=clip_norm) + else: + grad_clip = None + optim_model = [] + + # for static graph + if model_list is None: + optim = getattr(optimizer, optim_name)( + learning_rate=lr, grad_clip=grad_clip, + **optim_cfg)(model_list=optim_model) + return optim, lr + + # for dynamic graph + for i in range(len(model_list)): + if len(model_list[i].parameters()) == 0: + continue + if optim_scope == "all": + # optimizer for all + optim_model.append(model_list[i]) + else: + if optim_scope.endswith("Loss"): + # optimizer for loss + for m in model_list[i].sublayers(True): + if m.__class__.__name__ == optim_scope: + optim_model.append(m) + else: + # opmizer for module in model, such as backbone, neck, head... + if optim_scope == model_list[i].__class__.__name__: + optim_model.append(model_list[i]) + elif hasattr(model_list[i], optim_scope): + optim_model.append(getattr(model_list[i], optim_scope)) + + optim = getattr(optimizer, optim_name)( + learning_rate=lr, grad_clip=grad_clip, + **optim_cfg)(model_list=optim_model) + logger.debug("build optimizer ({}) for scope ({}) success..".format( + optim, optim_scope)) + optim_list.append(optim) + lr_list.append(lr) + return optim_list, lr_list diff --git a/ppcls/optimizer/learning_rate.py b/ppcls/optimizer/learning_rate.py index b59387dd935c805078ffdb435788373e07743807..1a4561133f948831b9ca0d69821a3394f092fae7 100644 --- a/ppcls/optimizer/learning_rate.py +++ b/ppcls/optimizer/learning_rate.py @@ -75,6 +75,23 @@ class Linear(object): return learning_rate +class Constant(LRScheduler): + """ + Constant learning rate + Args: + lr (float): The initial learning rate. It is a python float number. + last_epoch (int, optional): The index of last epoch. Can be set to restart training. Default: -1, means initial learning rate. + """ + + def __init__(self, learning_rate, last_epoch=-1, **kwargs): + self.learning_rate = learning_rate + self.last_epoch = last_epoch + super().__init__() + + def get_lr(self): + return self.learning_rate + + class Cosine(object): """ Cosine learning rate decay @@ -188,6 +205,7 @@ class Piecewise(object): The type of element in the list is python float. warmup_epoch(int): The epoch numbers for LinearWarmup. Default: 0. warmup_start_lr(float): Initial learning rate of warm up. Default: 0.0. + by_epoch(bool): Whether lr decay by epoch. Default: False. last_epoch (int, optional): The index of last epoch. Can be set to restart training. Default: -1, means initial learning rate. """ @@ -198,6 +216,7 @@ class Piecewise(object): epochs, warmup_epoch=0, warmup_start_lr=0.0, + by_epoch=False, last_epoch=-1, **kwargs): super().__init__() @@ -205,24 +224,41 @@ class Piecewise(object): msg = f"When using warm up, the value of \"Global.epochs\" must be greater than value of \"Optimizer.lr.warmup_epoch\". The value of \"Optimizer.lr.warmup_epoch\" has been set to {epochs}." logger.warning(msg) warmup_epoch = epochs - self.boundaries = [step_each_epoch * e for e in decay_epochs] + self.boundaries_steps = [step_each_epoch * e for e in decay_epochs] + self.boundaries_epoch = decay_epochs self.values = values self.last_epoch = last_epoch self.warmup_steps = round(warmup_epoch * step_each_epoch) + self.warmup_epoch = warmup_epoch self.warmup_start_lr = warmup_start_lr + self.by_epoch = by_epoch def __call__(self): - learning_rate = lr.PiecewiseDecay( - boundaries=self.boundaries, - values=self.values, - last_epoch=self.last_epoch) - if self.warmup_steps > 0: - learning_rate = lr.LinearWarmup( - learning_rate=learning_rate, - warmup_steps=self.warmup_steps, - start_lr=self.warmup_start_lr, - end_lr=self.values[0], + if self.by_epoch: + learning_rate = lr.PiecewiseDecay( + boundaries=self.boundaries_epoch, + values=self.values, + last_epoch=self.last_epoch) + if self.warmup_epoch > 0: + learning_rate = lr.LinearWarmup( + learning_rate=learning_rate, + warmup_steps=self.warmup_epoch, + start_lr=self.warmup_start_lr, + end_lr=self.values[0], + last_epoch=self.last_epoch) + else: + learning_rate = lr.PiecewiseDecay( + boundaries=self.boundaries_steps, + values=self.values, last_epoch=self.last_epoch) + if self.warmup_steps > 0: + learning_rate = lr.LinearWarmup( + learning_rate=learning_rate, + warmup_steps=self.warmup_steps, + start_lr=self.warmup_start_lr, + end_lr=self.values[0], + last_epoch=self.last_epoch) + setattr(learning_rate, "by_epoch", self.by_epoch) return learning_rate diff --git a/ppcls/optimizer/optimizer.py b/ppcls/optimizer/optimizer.py index 4422ea70d32a3ed1ce89c33ec806e2035aa25420..be6fa9f70ab8c50422e8f068b8e1c58a3b1c53e9 100644 --- a/ppcls/optimizer/optimizer.py +++ b/ppcls/optimizer/optimizer.py @@ -22,6 +22,51 @@ import paddle from ppcls.utils import logger +class SGD(object): + """ + Args: + learning_rate (float|Tensor|LearningRateDecay, optional): The learning rate used to update ``Parameter``. + It can be a float value, a ``Tensor`` with a float type or a LearningRateDecay. The default value is 0.001. + parameters (list|tuple, optional): List/Tuple of ``Tensor`` to update to minimize ``loss``. \ + This parameter is required in dygraph mode. \ + The default value is None in static mode, at this time all parameters will be updated. + weight_decay (float|WeightDecayRegularizer, optional): The strategy of regularization. \ + It canbe a float value as coeff of L2 regularization or \ + :ref:`api_fluid_regularizer_L1Decay`, :ref:`api_fluid_regularizer_L2Decay`. + If a parameter has set regularizer using :ref:`api_fluid_ParamAttr` already, \ + the regularization setting here in optimizer will be ignored for this parameter. \ + Otherwise, the regularization setting here in optimizer will take effect. \ + Default None, meaning there is no regularization. + grad_clip (GradientClipBase, optional): Gradient cliping strategy, it's an instance of + some derived class of ``GradientClipBase`` . There are three cliping strategies + ( :ref:`api_fluid_clip_GradientClipByGlobalNorm` , :ref:`api_fluid_clip_GradientClipByNorm` , + :ref:`api_fluid_clip_GradientClipByValue` ). Default None, meaning there is no gradient clipping. + name (str, optional): The default value is None. Normally there is no need for user + to set this property. + """ + + def __init__(self, + learning_rate=0.001, + weight_decay=None, + grad_clip=None, + name=None): + self.learning_rate = learning_rate + self.weight_decay = weight_decay + self.grad_clip = grad_clip + self.name = name + + def __call__(self, model_list): + # model_list is None in static graph + parameters = sum([m.parameters() for m in model_list], + []) if model_list else None + opt = optim.SGD(learning_rate=self.learning_rate, + parameters=parameters, + weight_decay=self.weight_decay, + grad_clip=self.grad_clip, + name=self.name) + return opt + + class Momentum(object): """ Simple Momentum optimizer with velocity state. diff --git a/ppcls/static/program.py b/ppcls/static/program.py index b3534a2cf24ec8b53969dd0458cab202c4e85b76..29107c9c1c1d8f571f0f8cf1cf0b7357ae3100ea 100644 --- a/ppcls/static/program.py +++ b/ppcls/static/program.py @@ -406,7 +406,7 @@ def run(dataloader, if "time" in key else str(metric_dict[key].value) for key in metric_dict ]) - ips_info = " ips: {:.5f} images/sec.".format( + ips_info = " ips: {:.5f} samples/sec.".format( batch_size / metric_dict["batch_time"].avg) fetchs_str += ips_info @@ -433,8 +433,8 @@ def run(dataloader, end_str = ' '.join([str(m.mean) for m in metric_dict.values()] + [metric_dict["batch_time"].total]) - ips_info = "ips: {:.5f} images/sec.".format(batch_size / - metric_dict["batch_time"].avg) + ips_info = "ips: {:.5f} samples/sec.".format(batch_size / + metric_dict["batch_time"].avg) if mode == 'eval': logger.info("END {:s} {:s} {:s}".format(mode, end_str, ips_info)) else: diff --git a/ppcls/static/train.py b/ppcls/static/train.py index 9c03598be96ec00e9d7198bc6d2231ed805dc801..eb803970b23954324e7870133b61fe4bb7122a0c 100644 --- a/ppcls/static/train.py +++ b/ppcls/static/train.py @@ -71,7 +71,7 @@ def main(args): log_file = os.path.join(global_config['output_dir'], config["Arch"]["name"], f"{mode}.log") - init_logger(name='root', log_file=log_file) + init_logger(log_file=log_file) print_config(config) if global_config.get("is_distributed", True): @@ -91,9 +91,10 @@ def main(args): use_xpu = global_config.get("use_xpu", False) use_npu = global_config.get("use_npu", False) + use_mlu = global_config.get("use_mlu", False) assert ( - use_gpu and use_xpu and use_npu - ) is not True, "gpu, xpu and npu can not be true in the same time in static mode!" + use_gpu and use_xpu and use_npu and use_mlu + ) is not True, "gpu, xpu, npu and mlu can not be true in the same time in static mode!" if use_gpu: device = paddle.set_device('gpu') @@ -101,6 +102,8 @@ def main(args): device = paddle.set_device('xpu') elif use_npu: device = paddle.set_device('npu') + elif use_mlu: + device = paddle.set_device('mlu') else: device = paddle.set_device('cpu') @@ -158,12 +161,22 @@ def main(args): # load pretrained models or checkpoints init_model(global_config, train_prog, exe) - if 'AMP' in config and config.AMP.get("level", "O1") == "O2": + if 'AMP' in config: + if config["AMP"].get("level", "O1").upper() == "O2": + use_fp16_test = True + msg = "Only support FP16 evaluation when AMP O2 is enabled." + logger.warning(msg) + elif "use_fp16_test" in config["AMP"]: + use_fp16_test = config["AMP"].get["use_fp16_test"] + else: + use_fp16_test = False + optimizer.amp_init( device, scope=paddle.static.global_scope(), test_program=eval_prog - if global_config["eval_during_train"] else None) + if global_config["eval_during_train"] else None, + use_fp16_test=use_fp16_test) if not global_config.get("is_distributed", True): compiled_train_prog = program.compile( @@ -179,7 +192,7 @@ def main(args): program.run(train_dataloader, exe, compiled_train_prog, train_feeds, train_fetchs, epoch_id, 'train', config, vdl_writer, lr_scheduler, args.profiler_options) - # 2. evaate with eval dataset + # 2. evaluate with eval dataset if global_config["eval_during_train"] and epoch_id % global_config[ "eval_interval"] == 0: top1_acc = program.run(eval_dataloader, exe, compiled_eval_prog, diff --git a/ppcls/utils/config.py b/ppcls/utils/config.py index e3277c480943cdfb7ce49f4f3ea7bbd160c34ebb..1c2f5fdb8b40a19e7373e6298e3f6b6169e9e8b4 100644 --- a/ppcls/utils/config.py +++ b/ppcls/utils/config.py @@ -137,9 +137,12 @@ def override(dl, ks, v): if len(ks) == 1: # assert ks[0] in dl, ('{} is not exist in {}'.format(ks[0], dl)) if not ks[0] in dl: - print('A new filed ({}) detected!'.format(ks[0], dl)) + print('A new field ({}) detected!'.format(ks[0], dl)) dl[ks[0]] = str2num(v) else: + if ks[0] not in dl.keys(): + dl[ks[0]] = {} + print("A new Series field ({}) detected!".format(ks[0], dl)) override(dl[ks[0]], ks[1:], v) diff --git a/ppcls/utils/gallery2fc.py b/ppcls/utils/gallery2fc.py index 67b08529ebb37a8a147aa81f3ebb80dc13296624..e654522cf5d70eaa162581dd684fb4131d4f91c3 100644 --- a/ppcls/utils/gallery2fc.py +++ b/ppcls/utils/gallery2fc.py @@ -12,10 +12,14 @@ # See the License for the specific language governing permissions and # limitations under the License. -import os import paddle import cv2 +import os +import sys +__dir__ = os.path.dirname(os.path.abspath(__file__)) +sys.path.append(os.path.abspath(os.path.join(__dir__, '../../'))) + from ppcls.arch import build_model from ppcls.utils.config import parse_config, parse_args from ppcls.utils.save_load import load_dygraph_pretrain @@ -51,12 +55,17 @@ class GalleryLayer(paddle.nn.Layer): self.gallery_images.append(image_file) gallery_docs.append(ori_line.strip()) gallery_labels.append(line[1].strip()) - self.gallery_layer = paddle.nn.Linear(embedding_size, len(self.gallery_images), bias_attr=False) + self.gallery_layer = paddle.nn.Linear( + embedding_size, len(self.gallery_images), bias_attr=False) self.gallery_layer.skip_quant = True output_label_str = "" for i, label_i in enumerate(gallery_labels): output_label_str += "{} {}\n".format(i, label_i) output_path = configs["Global"]["save_inference_dir"] + "_label.txt" + + save_dir = os.path.dirname(configs["Global"]["save_inference_dir"]) + if not os.path.exists(save_dir): + os.makedirs(save_dir) with open(output_path, "w") as f: f.write(output_label_str) @@ -71,19 +80,23 @@ class GalleryLayer(paddle.nn.Layer): embedding_size = self.configs["Arch"]["Head"]["embedding_size"] batch_index = 0 input_tensor = paddle.zeros(self.image_shape) - gallery_feature = paddle.zeros((len(self.gallery_images), embedding_size)) + gallery_feature = paddle.zeros( + (len(self.gallery_images), embedding_size)) for i, image_path in enumerate(self.gallery_images): image = cv2.imread(image_path)[:, :, ::-1] for op in preprocess_ops: image = op(image) input_tensor[batch_index] = image batch_index += 1 - if batch_index == self.batch_size or i == len(self.gallery_images) - 1: + if batch_index == self.batch_size or i == len( + self.gallery_images) - 1: batch_feature = feature_extractor(input_tensor)["features"] for j in range(batch_index): feature = batch_feature[j] - norm_feature = paddle.nn.functional.normalize(feature, axis=0) + norm_feature = paddle.nn.functional.normalize( + feature, axis=0) gallery_feature[i - batch_index + j + 1] = norm_feature + batch_index = 0 self.gallery_layer.set_state_dict({"_layer.weight": gallery_feature.T}) diff --git a/ppcls/utils/logger.py b/ppcls/utils/logger.py index d4faaa9b0cd77b6038b40277cc62b4c6981f2da4..bc8de364091e9b56dafdcffa4475f7f225306e1b 100644 --- a/ppcls/utils/logger.py +++ b/ppcls/utils/logger.py @@ -22,7 +22,7 @@ import paddle.distributed as dist _logger = None -def init_logger(name='root', log_file=None, log_level=logging.INFO): +def init_logger(name='ppcls', log_file=None, log_level=logging.INFO): """Initialize and get a logger by name. If the logger has not been initialized, this method will initialize the logger by adding one or two handlers, otherwise the initialized logger will @@ -59,6 +59,7 @@ def init_logger(name='root', log_file=None, log_level=logging.INFO): _logger.setLevel(log_level) else: _logger.setLevel(logging.ERROR) + _logger.propagate = False def log_at_trainer0(log): diff --git a/ppcls/utils/save_load.py b/ppcls/utils/save_load.py index 625a2848339e47e688a263631cfe4a5b9e6c691b..4e27f12c1d4830f2f16580bfa976cf3ace78d934 100644 --- a/ppcls/utils/save_load.py +++ b/ppcls/utils/save_load.py @@ -18,9 +18,6 @@ from __future__ import print_function import errno import os -import re -import shutil -import tempfile import paddle from ppcls.utils import logger @@ -47,10 +44,15 @@ def _mkdir_if_not_exist(path): def load_dygraph_pretrain(model, path=None): if not (os.path.isdir(path) or os.path.exists(path + '.pdparams')): - raise ValueError("Model pretrain path {} does not " + raise ValueError("Model pretrain path {}.pdparams does not " "exists.".format(path)) param_state_dict = paddle.load(path + ".pdparams") - model.set_dict(param_state_dict) + if isinstance(model, list): + for m in model: + if hasattr(m, 'set_dict'): + m.set_dict(param_state_dict) + else: + model.set_dict(param_state_dict) return @@ -85,7 +87,7 @@ def load_distillation_model(model, pretrained_model): pretrained_model)) -def init_model(config, net, optimizer=None): +def init_model(config, net, optimizer=None, loss: paddle.nn.Layer=None): """ load model from checkpoint or pretrained_model """ @@ -95,11 +97,15 @@ def init_model(config, net, optimizer=None): "Given dir {}.pdparams not exist.".format(checkpoints) assert os.path.exists(checkpoints + ".pdopt"), \ "Given dir {}.pdopt not exist.".format(checkpoints) - para_dict = paddle.load(checkpoints + ".pdparams") + # load state dict opti_dict = paddle.load(checkpoints + ".pdopt") + para_dict = paddle.load(checkpoints + ".pdparams") metric_dict = paddle.load(checkpoints + ".pdstates") - net.set_dict(para_dict) - optimizer.set_state_dict(opti_dict) + # set state dict + net.set_state_dict(para_dict) + loss.set_state_dict(para_dict) + for i in range(len(optimizer)): + optimizer[i].set_state_dict(opti_dict) logger.info("Finish load checkpoints from {}".format(checkpoints)) return metric_dict @@ -110,9 +116,8 @@ def init_model(config, net, optimizer=None): load_distillation_model(net, pretrained_model) else: # common load load_dygraph_pretrain(net, path=pretrained_model) - logger.info( - logger.coloring("Finish load pretrained model from {}".format( - pretrained_model), "HEADER")) + logger.info("Finish load pretrained model from {}".format( + pretrained_model)) def save_model(net, @@ -120,7 +125,8 @@ def save_model(net, metric_info, model_path, model_name="", - prefix='ppcls'): + prefix='ppcls', + loss: paddle.nn.Layer=None): """ save model to the target path """ @@ -130,7 +136,14 @@ def save_model(net, _mkdir_if_not_exist(model_path) model_path = os.path.join(model_path, prefix) - paddle.save(net.state_dict(), model_path + ".pdparams") - paddle.save(optimizer.state_dict(), model_path + ".pdopt") + params_state_dict = net.state_dict() + loss_state_dict = loss.state_dict() + keys_inter = set(params_state_dict.keys()) & set(loss_state_dict.keys()) + assert len(keys_inter) == 0, \ + f"keys in model and loss state_dict must be unique, but got intersection {keys_inter}" + params_state_dict.update(loss_state_dict) + + paddle.save(params_state_dict, model_path + ".pdparams") + paddle.save([opt.state_dict() for opt in optimizer], model_path + ".pdopt") paddle.save(metric_info, model_path + ".pdstates") logger.info("Already save model in {}".format(model_path)) diff --git a/setup.py b/setup.py index a17e77d4fb2d17f27e4225d9f90b97cd4320c6fa..57045d31903917fdb8634887a1f6e7207871ead5 100644 --- a/setup.py +++ b/setup.py @@ -20,7 +20,9 @@ with open('requirements.txt', encoding="utf-8-sig") as f: def readme(): - with open('docs/en/whl_en.md', encoding="utf-8-sig") as f: + with open( + 'docs/en/inference_deployment/whl_deploy_en.md', + encoding="utf-8-sig") as f: README = f.read() return README diff --git a/test_tipc/benchmark_train.sh b/test_tipc/benchmark_train.sh new file mode 100644 index 0000000000000000000000000000000000000000..793b89476fb829034687b442c517546f5d8a4cfc --- /dev/null +++ b/test_tipc/benchmark_train.sh @@ -0,0 +1,268 @@ +#!/bin/bash +source test_tipc/common_func.sh + +# set env +python=python +export str_tmp=$(echo `pip list|grep paddlepaddle-gpu|awk -F ' ' '{print $2}'`) +export frame_version=${str_tmp%%.post*} +export frame_commit=$(echo `${python} -c "import paddle;print(paddle.version.commit)"`) + +# run benchmark sh +# Usage: +# bash run_benchmark_train.sh config.txt params +# or +# bash run_benchmark_train.sh config.txt + +function func_parser_params(){ + strs=$1 + IFS="=" + array=(${strs}) + tmp=${array[1]} + echo ${tmp} +} + +function func_sed_params(){ + filename=$1 + line=$2 + param_value=$3 + params=`sed -n "${line}p" $filename` + IFS=":" + array=(${params}) + key=${array[0]} + value=${array[1]} + new_params="${key}:${param_value}" + IFS=";" + cmd="sed -i '${line}s/.*/${new_params}/' '${filename}'" + eval $cmd +} + +function set_gpu_id(){ + string=$1 + _str=${string:1:6} + IFS="C" + arr=(${_str}) + M=${arr[0]} + P=${arr[1]} + gn=`expr $P - 1` + gpu_num=`expr $gn / $M` + seq=`seq -s "," 0 $gpu_num` + echo $seq +} + +function get_repo_name(){ + IFS=";" + cur_dir=$(pwd) + IFS="/" + arr=(${cur_dir}) + echo ${arr[-1]} +} + +FILENAME=$1 +# copy FILENAME as new +new_filename="./test_tipc/benchmark_train.txt" +cmd=`yes|cp $FILENAME $new_filename` +FILENAME=$new_filename +# MODE must be one of ['benchmark_train'] +MODE=$2 +PARAMS=$3 +model_type=$4 +# bash test_tipc/benchmark_train.sh test_tipc/configs/det_mv3_db_v2_0/train_benchmark.txt benchmark_train +# bash test_tipc/benchmark_train.sh test_tipc/configs/det_mv3_db_v2_0/train_benchmark.txt benchmark_train dynamicTostatic +# bash test_tipc/benchmark_train.sh test_tipc/configs/det_mv3_db_v2_0/train_benchmark.txt benchmark_train dynamic_bs8_null_DP_N1C1 +# bash test_tipc/benchmark_train.sh test_tipc/configs/det_mv3_db_v2_0/train_benchmark.txt benchmark_train dynamicTostatic_bs8_null_DP_N1C1 + +IFS=$'\n' +# parser params from train_benchmark.txt +sed -i 's/ -o DataLoader.Train.sampler.shuffle=False//g' $FILENAME +sed -i 's/ -o DataLoader.Train.loader.num_workers=0/ -o Global.print_batch_step=1/g' $FILENAME +sed -i 's/-o DataLoader.Train.loader.use_shared_memory=False/ -o Global.eval_during_train=False/g' $FILENAME +dataline=`cat $FILENAME` +# parser params +IFS=$'\n' +lines=(${dataline}) +model_name=$(func_parser_value "${lines[1]}") + +# 获取benchmark_params所在的行数 +line_num=`grep -n "train_benchmark_params" $FILENAME | cut -d ":" -f 1` +# for train log parser +batch_size=$(func_parser_value "${lines[line_num]}") +line_num=`expr $line_num + 1` +fp_items=$(func_parser_value "${lines[line_num]}") +line_num=`expr $line_num + 1` +epoch=$(func_parser_value "${lines[line_num]}") + +line_num=`expr $line_num + 1` +profile_option_key=$(func_parser_key "${lines[line_num]}") +profile_option_params=$(func_parser_value "${lines[line_num]}") +profile_option="${profile_option_key}:${profile_option_params}" + +line_num=`expr $line_num + 1` +flags_value=$(func_parser_value "${lines[line_num]}") +# set flags +IFS=";" +flags_list=(${flags_value}) +for _flag in ${flags_list[*]}; do + cmd="export ${_flag}" + eval $cmd +done + +# set log_name +repo_name=$(get_repo_name ) +SAVE_LOG=${BENCHMARK_LOG_DIR:-$(pwd)} # */benchmark_log +mkdir -p "${SAVE_LOG}/benchmark_log/" +status_log="${SAVE_LOG}/benchmark_log/results.log" + +# The number of lines in which train params can be replaced. +line_python=3 +line_gpuid=4 +line_precision=6 +line_epoch=7 +line_batchsize=9 +line_profile=13 +line_eval_py=24 +line_export_py=30 +line_norm_train=16 +line_pact_train=17 +line_fgpm_train=18 + +func_sed_params "$FILENAME" "${line_eval_py}" "null" +func_sed_params "$FILENAME" "${line_export_py}" "null" +func_sed_params "$FILENAME" "${line_python}" "$python" +func_sed_params "$FILENAME" "${line_pact_train}" "null" +func_sed_params "$FILENAME" "${line_fgpm_train}" "null" + +# if params +if [ ! -n "$PARAMS" ] ;then + # PARAMS input is not a word. + IFS="|" + batch_size_list=(${batch_size}) + fp_items_list=(${fp_items}) + device_num_list=(N1C4) + run_mode="DP" +else + # parser params from input: modeltype_bs${bs_item}_${fp_item}_${run_mode}_${device_num} + IFS="_" + params_list=(${PARAMS}) + model_type=${params_list[0]} + batch_size=${params_list[1]} + batch_size=`echo ${batch_size} | tr -cd "[0-9]" ` + precision=${params_list[2]} + run_mode=${params_list[3]} + device_num=${params_list[4]} + IFS=";" + + if [ ${precision} = "null" ];then + precision="fp32" + fi + + fp_items_list=($precision) + batch_size_list=($batch_size) + device_num_list=($device_num) +fi + +# for log name +to_static="" +# parse "to_static" options and modify trainer into "to_static_trainer" +if [ $model_type = "dynamicTostatic" ] ;then + to_static="d2sT_" + sed -i 's/trainer:norm_train/trainer:to_static_train/g' $FILENAME +fi + + +IFS="|" +for batch_size in ${batch_size_list[*]}; do + for precision in ${fp_items_list[*]}; do + for device_num in ${device_num_list[*]}; do + # sed batchsize and precision + func_sed_params "$FILENAME" "${line_precision}" "$precision" + func_sed_params "$FILENAME" "${line_batchsize}" "$batch_size" + func_sed_params "$FILENAME" "${line_epoch}" "$epoch" + gpu_id=$(set_gpu_id $device_num) + + if [ ${#gpu_id} -le 1 ];then + log_path="$SAVE_LOG/profiling_log" + mkdir -p $log_path + log_name="${repo_name}_${model_name}_bs${batch_size}_${precision}_${run_mode}_${device_num}_${to_static}profiling" + func_sed_params "$FILENAME" "${line_gpuid}" "0" # sed used gpu_id + # set profile_option params + tmp=`sed -i "${line_profile}s/.*/${profile_option}/" "${FILENAME}"` + + # run test_train_inference_python.sh + cmd="bash test_tipc/test_train_inference_python.sh ${FILENAME} benchmark_train > ${log_path}/${log_name} 2>&1 " + echo $cmd + eval $cmd + eval "cat ${log_path}/${log_name}" + + # without profile + log_path="$SAVE_LOG/train_log" + speed_log_path="$SAVE_LOG/index" + mkdir -p $log_path + mkdir -p $speed_log_path + log_name="${repo_name}_${model_name}_bs${batch_size}_${precision}_${run_mode}_${device_num}_${to_static}log" + speed_log_name="${repo_name}_${model_name}_bs${batch_size}_${precision}_${run_mode}_${device_num}_${to_static}speed" + func_sed_params "$FILENAME" "${line_profile}" "null" # sed profile_id as null + cmd="bash test_tipc/test_train_inference_python.sh ${FILENAME} benchmark_train > ${log_path}/${log_name} 2>&1 " + echo $cmd + job_bt=`date '+%Y%m%d%H%M%S'` + eval $cmd + job_et=`date '+%Y%m%d%H%M%S'` + export model_run_time=$((${job_et}-${job_bt})) + eval "cat ${log_path}/${log_name}" + + # parser log + _model_name="${model_name}_bs${batch_size}_${precision}_${run_mode}" + cmd="${python} ${BENCHMARK_ROOT}/scripts/analysis.py --filename ${log_path}/${log_name} \ + --speed_log_file '${speed_log_path}/${speed_log_name}' \ + --model_name ${_model_name} \ + --base_batch_size ${batch_size} \ + --run_mode ${run_mode} \ + --fp_item ${precision} \ + --keyword ips: \ + --skip_steps 2 \ + --device_num ${device_num} \ + --speed_unit samples/s \ + --convergence_key loss: " + echo $cmd + eval $cmd + last_status=${PIPESTATUS[0]} + status_check $last_status "${cmd}" "${status_log}" + else + IFS=";" + unset_env=`unset CUDA_VISIBLE_DEVICES` + log_path="$SAVE_LOG/train_log" + speed_log_path="$SAVE_LOG/index" + mkdir -p $log_path + mkdir -p $speed_log_path + log_name="${repo_name}_${model_name}_bs${batch_size}_${precision}_${run_mode}_${device_num}_${to_static}log" + speed_log_name="${repo_name}_${model_name}_bs${batch_size}_${precision}_${run_mode}_${device_num}_${to_static}speed" + func_sed_params "$FILENAME" "${line_gpuid}" "$gpu_id" # sed used gpu_id + func_sed_params "$FILENAME" "${line_profile}" "null" # sed --profile_option as null + cmd="bash test_tipc/test_train_inference_python.sh ${FILENAME} benchmark_train > ${log_path}/${log_name} 2>&1 " + echo $cmd + job_bt=`date '+%Y%m%d%H%M%S'` + eval $cmd + job_et=`date '+%Y%m%d%H%M%S'` + export model_run_time=$((${job_et}-${job_bt})) + eval "cat ${log_path}/${log_name}" + # parser log + _model_name="${model_name}_bs${batch_size}_${precision}_${run_mode}" + + cmd="${python} ${BENCHMARK_ROOT}/scripts/analysis.py --filename ${log_path}/${log_name} \ + --speed_log_file '${speed_log_path}/${speed_log_name}' \ + --model_name ${_model_name} \ + --base_batch_size ${batch_size} \ + --run_mode ${run_mode} \ + --fp_item ${precision} \ + --keyword ips: \ + --skip_steps 2 \ + --device_num ${device_num} \ + --speed_unit images/s \ + --convergence_key loss: " + echo $cmd + eval $cmd + last_status=${PIPESTATUS[0]} + status_check $last_status "${cmd}" "${status_log}" + fi + done + done +done diff --git a/test_tipc/common_func.sh b/test_tipc/common_func.sh index 3f0fa66b77ff50b23b1e83dea506580f549f8ecf..63fa1014487ce43405896ddf97f5d2aae0344489 100644 --- a/test_tipc/common_func.sh +++ b/test_tipc/common_func.sh @@ -16,6 +16,14 @@ function func_parser_value(){ echo ${tmp} } +function func_parser_value_lite(){ + strs=$1 + IFS=$2 + array=(${strs}) + tmp=${array[1]} + echo ${tmp} +} + function func_set_params(){ key=$1 value=$2 diff --git a/test_tipc/config/AlexNet/AlexNet_train_amp_infer_python.txt b/test_tipc/config/AlexNet/AlexNet_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..c0cf046186576e77776c71699611bb541ddf1aaf --- /dev/null +++ b/test_tipc/config/AlexNet/AlexNet_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:AlexNet +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/AlexNet/AlexNet.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/AlexNet/AlexNet.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/AlexNet/AlexNet.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/AlexNet_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/AlexNet/AlexNet_train_infer_python.txt b/test_tipc/config/AlexNet/AlexNet_train_infer_python.txt index 7662f5d92832a6d880e7b55a4beb6190078c0ce7..56e19cd417967b51f14f20fad304a0be39f9ec7b 100644 --- a/test_tipc/config/AlexNet/AlexNet_train_infer_python.txt +++ b/test_tipc/config/AlexNet/AlexNet_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/CSPNet/CSPDarkNet53_train_amp_infer_python.txt b/test_tipc/config/CSPNet/CSPDarkNet53_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..534cb09f64aa706c7b1d6a225fac7fcf74d8fdef --- /dev/null +++ b/test_tipc/config/CSPNet/CSPDarkNet53_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:CSPDarkNet53 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/CSPNet/CSPDarkNet53.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSPNet/CSPDarkNet53.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSPNet/CSPDarkNet53.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSPDarkNet53_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=288 -o PreProcess.transform_ops.1.CropImage.size=256 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/CSPNet/CSPDarkNet53_train_infer_python.txt b/test_tipc/config/CSPNet/CSPDarkNet53_train_infer_python.txt index cd5ec06d8ae0331a3626929834cd920f9ca8b987..bf82ef353d6446d639341d928df204c820fd9e3f 100644 --- a/test_tipc/config/CSPNet/CSPDarkNet53_train_infer_python.txt +++ b/test_tipc/config/CSPNet/CSPDarkNet53_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,256,256]}] \ No newline at end of file diff --git a/test_tipc/config/CSWinTransformer/CSWinTransformer_base_224_train_infer_python.txt b/test_tipc/config/CSWinTransformer/CSWinTransformer_base_224_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3de6add85078c13a323e0f62a8d805880f189c6c --- /dev/null +++ b/test_tipc/config/CSWinTransformer/CSWinTransformer_base_224_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:CSWinTransformer_base_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=248 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/CSWinTransformer/CSWinTransformer_base_384_train_infer_python.txt b/test_tipc/config/CSWinTransformer/CSWinTransformer_base_384_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3a913e253ee31e04d5c3d86b3be34aeb0e777949 --- /dev/null +++ b/test_tipc/config/CSWinTransformer/CSWinTransformer_base_384_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:CSWinTransformer_base_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_base_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_base_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] diff --git a/test_tipc/config/CSWinTransformer/CSWinTransformer_large_224_train_infer_python.txt b/test_tipc/config/CSWinTransformer/CSWinTransformer_large_224_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..d376d4cba38f52522fb7ecc9a93e38bef1b7b710 --- /dev/null +++ b/test_tipc/config/CSWinTransformer/CSWinTransformer_large_224_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:CSWinTransformer_large_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=248 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/CSWinTransformer/CSWinTransformer_large_384_train_infer_python.txt b/test_tipc/config/CSWinTransformer/CSWinTransformer_large_384_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..781fefe58dbe18b1bbf89026b59b95c995ecf0c5 --- /dev/null +++ b/test_tipc/config/CSWinTransformer/CSWinTransformer_large_384_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:CSWinTransformer_large_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:4 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_large_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_large_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] diff --git a/test_tipc/config/CSWinTransformer/CSWinTransformer_small_224_train_infer_python.txt b/test_tipc/config/CSWinTransformer/CSWinTransformer_small_224_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..8057bb84b766e0f618b93a85f3f69f82fc8fa712 --- /dev/null +++ b/test_tipc/config/CSWinTransformer/CSWinTransformer_small_224_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:CSWinTransformer_small_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_small_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_small_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_small_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_small_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=248 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/CSWinTransformer/CSWinTransformer_tiny_224_train_infer_python.txt b/test_tipc/config/CSWinTransformer/CSWinTransformer_tiny_224_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..55afdbc96cd7a6a82b5350a7a2b3e4da804d6ba0 --- /dev/null +++ b/test_tipc/config/CSWinTransformer/CSWinTransformer_tiny_224_train_infer_python.txt @@ -0,0 +1,60 @@ +===========================train_params=========================== +model_name:CSWinTransformer_tiny_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_tiny_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o Global.print_batch_step=1 +pact_train:null +fpgm_train:null +distill_train:null +to_static_train:-o Global.to_static=True +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_tiny_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/CSWinTransformer/CSWinTransformer_tiny_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/CSWinTransformer_tiny_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=248 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================train_benchmark_params========================== +batch_size:128 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/DLA/DLA102_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA102_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..bdc23ffa35fde038393a59546e3607573ee27ff8 --- /dev/null +++ b/test_tipc/config/DLA/DLA102_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA102 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA102.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA102.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA102.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA102_train_infer_python.txt b/test_tipc/config/DLA/DLA102_train_infer_python.txt index 0a85e2a3541e7d9aa945e1952eca90ebd7f79200..e4c0edf0b229edb205531ce0c1437d6e43fdf322 100644 --- a/test_tipc/config/DLA/DLA102_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA102_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA102x2_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA102x2_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..2552690f31120311edb7dee5c4b334bc131d01d2 --- /dev/null +++ b/test_tipc/config/DLA/DLA102x2_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA102x2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA102x2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA102x2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA102x2.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x2_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA102x2_train_infer_python.txt b/test_tipc/config/DLA/DLA102x2_train_infer_python.txt index fff0ee1c562db6bd049a3e3d68a5dd5e5a5ff17b..cdbc367c8dbd543e60e964fc748fa98fb6861dd2 100644 --- a/test_tipc/config/DLA/DLA102x2_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA102x2_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA102x_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA102x_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f1c19957bde3d68021ae83056a7ef5d429bdade4 --- /dev/null +++ b/test_tipc/config/DLA/DLA102x_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA102x +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA102x.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA102x.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA102x.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA102x_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA102x_train_infer_python.txt b/test_tipc/config/DLA/DLA102x_train_infer_python.txt index db7af68caed4bd4b979ea2182de15dc2aa3285d0..f0316cb2eb6416b3a17112e23a74bd35694836f4 100644 --- a/test_tipc/config/DLA/DLA102x_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA102x_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA169_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA169_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..8ba21d09a359c1e86faf4557762edeaf4d911274 --- /dev/null +++ b/test_tipc/config/DLA/DLA169_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA169 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA169.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA169.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA169.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA169_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA169_train_infer_python.txt b/test_tipc/config/DLA/DLA169_train_infer_python.txt index 0abc0e21828c834051927f9c9812a7c38f2ea842..f3f23a8ec1db951aa73a1604fd3456120c30bdf0 100644 --- a/test_tipc/config/DLA/DLA169_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA169_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA34_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA34_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..5e63e494dcd60d83d21bc9f2050fe3e7f2d545de --- /dev/null +++ b/test_tipc/config/DLA/DLA34_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA34 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA34.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA34.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA34.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA34_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA34_train_infer_python.txt b/test_tipc/config/DLA/DLA34_train_infer_python.txt index fa064056fd990a0402d559e85142f16274f02968..8360f0c05e59784322abdefaffb186ecacd39c08 100644 --- a/test_tipc/config/DLA/DLA34_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA34_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA46_c_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA46_c_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a9ec71f73cb1e5c60d5b4a77da4be21f8f69e305 --- /dev/null +++ b/test_tipc/config/DLA/DLA46_c_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA46_c +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA46_c.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA46_c.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA46_c.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA46_c_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA46_c_train_infer_python.txt b/test_tipc/config/DLA/DLA46_c_train_infer_python.txt index 8503452917db42672e2113f1fda4400cc57a824e..1ca4c2725899c3b25bce3e5ddad7e88891da64e2 100644 --- a/test_tipc/config/DLA/DLA46_c_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA46_c_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA46x_c_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA46x_c_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0325c27822419e75cb4c882749383528ac76bed9 --- /dev/null +++ b/test_tipc/config/DLA/DLA46x_c_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA46x_c +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA46x_c.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA46x_c.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA46x_c.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA46x_c_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA46x_c_train_infer_python.txt b/test_tipc/config/DLA/DLA46x_c_train_infer_python.txt index ee6c58ad8151110ebf9fa4577d0b10b3c3a5c93e..3a3687523070e0b04be664305936904e5429011c 100644 --- a/test_tipc/config/DLA/DLA46x_c_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA46x_c_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA60_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA60_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ac7b55c22f25c1a14c7f5a61350b86611c776922 --- /dev/null +++ b/test_tipc/config/DLA/DLA60_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA60 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA60.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA60.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA60.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA60_train_infer_python.txt b/test_tipc/config/DLA/DLA60_train_infer_python.txt index e14fcabeaaad10b3b18e4d0220233af36b102a8d..8bb2fbd982fa700a62370c79004a1160686e36c4 100644 --- a/test_tipc/config/DLA/DLA60_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA60_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA60x_c_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA60x_c_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..7b66c715e6c95cdd30223f4efb61a846837d1e25 --- /dev/null +++ b/test_tipc/config/DLA/DLA60x_c_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA60x_c +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA60x_c.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA60x_c.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA60x_c.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_c_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA60x_c_train_infer_python.txt b/test_tipc/config/DLA/DLA60x_c_train_infer_python.txt index cd45206c39bf36ab63fe05d237dd2ba37f1a347f..a9e3a1cfaa5382ec47d387174095f394cf2fcea6 100644 --- a/test_tipc/config/DLA/DLA60x_c_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA60x_c_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DLA/DLA60x_train_amp_infer_python.txt b/test_tipc/config/DLA/DLA60x_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4f775c9431e67ac4494f20771f8ba7800c327c3f --- /dev/null +++ b/test_tipc/config/DLA/DLA60x_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DLA60x +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DLA/DLA60x.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DLA/DLA60x.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DLA/DLA60x.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DLA60x_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DLA/DLA60x_train_infer_python.txt b/test_tipc/config/DLA/DLA60x_train_infer_python.txt index 51b7e77c8b8acd97fd999a1bfb997bbd90c12f53..274f5a4253e013a8b67671fac01e636ba83af97f 100644 --- a/test_tipc/config/DLA/DLA60x_train_infer_python.txt +++ b/test_tipc/config/DLA/DLA60x_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DPN/DPN107_train_amp_infer_python.txt b/test_tipc/config/DPN/DPN107_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..41102a8b14fe0c3e9c2923e88f19e193f68df6c2 --- /dev/null +++ b/test_tipc/config/DPN/DPN107_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN107 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN107.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN107.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN107.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN107_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DPN/DPN107_train_infer_python.txt b/test_tipc/config/DPN/DPN107_train_infer_python.txt index 646bce35aa0720a27919b5a323ce608f5f351991..48b5cdbbb134d662fa40ad4e1c20236220786252 100644 --- a/test_tipc/config/DPN/DPN107_train_infer_python.txt +++ b/test_tipc/config/DPN/DPN107_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DPN/DPN131_train_amp_infer_python.txt b/test_tipc/config/DPN/DPN131_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1423a3e324df402f5e4cebb584d4616dd047e2ae --- /dev/null +++ b/test_tipc/config/DPN/DPN131_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN131 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN131.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN131.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN131.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN131_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DPN/DPN131_train_infer_python.txt b/test_tipc/config/DPN/DPN131_train_infer_python.txt index 1aaea6a419146b6e886154bc15c8ba6ea8450259..bbb7899c1c8c340d93654d6424b01bbde4a1ee21 100644 --- a/test_tipc/config/DPN/DPN131_train_infer_python.txt +++ b/test_tipc/config/DPN/DPN131_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DPN/DPN68_train_amp_infer_python.txt b/test_tipc/config/DPN/DPN68_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1a8ae3d6168c6f34023c266ef6d413bc9db683c6 --- /dev/null +++ b/test_tipc/config/DPN/DPN68_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN68 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN68.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN68.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN68.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN68_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DPN/DPN68_train_infer_python.txt b/test_tipc/config/DPN/DPN68_train_infer_python.txt index d26e5e3fd47dcbc44a89121b3af2843bc294b688..91a24e8d3cf3f170d88743eb58cd7794d2c07452 100644 --- a/test_tipc/config/DPN/DPN68_train_infer_python.txt +++ b/test_tipc/config/DPN/DPN68_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DPN/DPN92_train_amp_infer_python.txt b/test_tipc/config/DPN/DPN92_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..bccd5ffd03d320ca11d4105ce2055a0bf9177565 --- /dev/null +++ b/test_tipc/config/DPN/DPN92_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN92 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN92.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN92.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN92.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN92_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DPN/DPN92_train_infer_python.txt b/test_tipc/config/DPN/DPN92_train_infer_python.txt index 3d0d0778dece0f7751c81a14f17a9dcb15413bea..d160a970134085c6e40ff60eedd43bb2dca4dbbc 100644 --- a/test_tipc/config/DPN/DPN92_train_infer_python.txt +++ b/test_tipc/config/DPN/DPN92_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DPN/DPN98_train_amp_infer_python.txt b/test_tipc/config/DPN/DPN98_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0e488de5e3d8502b53b8b0148e7769a21df0b4ad --- /dev/null +++ b/test_tipc/config/DPN/DPN98_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DPN98 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DPN/DPN98.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DPN/DPN98.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DPN/DPN98.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DPN98_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DPN/DPN98_train_infer_python.txt b/test_tipc/config/DPN/DPN98_train_infer_python.txt index e4cee43c8434b9b8c473ae31fea0b7a645a0ce7d..3dc6f133c8be5a63047cfa65e7ba13e9a4051eeb 100644 --- a/test_tipc/config/DPN/DPN98_train_infer_python.txt +++ b/test_tipc/config/DPN/DPN98_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DarkNet/DarkNet53_train_amp_infer_python.txt b/test_tipc/config/DarkNet/DarkNet53_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..907c8369786e4cfe29526d2a2418bf055e20a6c8 --- /dev/null +++ b/test_tipc/config/DarkNet/DarkNet53_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DarkNet53 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DarkNet/DarkNet53.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DarkNet53_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=292 -o PreProcess.transform_ops.1.CropImage.size=256 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DarkNet/DarkNet53_train_infer_python.txt b/test_tipc/config/DarkNet/DarkNet53_train_infer_python.txt index b10918ea41fd99f2b09b36cfe94ee38e9d3cfc62..31e38095e8880c1cd6689622b6d0a2aa24fac080 100644 --- a/test_tipc/config/DarkNet/DarkNet53_train_infer_python.txt +++ b/test_tipc/config/DarkNet/DarkNet53_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,256,256]}] \ No newline at end of file diff --git a/test_tipc/config/DeiT/DeiT_base_patch16_224_train_amp_infer_python.txt b/test_tipc/config/DeiT/DeiT_base_patch16_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1c52e7bb689f81d40b90c2f3c274fbe6f538c4e3 --- /dev/null +++ b/test_tipc/config/DeiT/DeiT_base_patch16_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_base_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DeiT/DeiT_base_patch16_224_train_infer_python.txt b/test_tipc/config/DeiT/DeiT_base_patch16_224_train_infer_python.txt index d10fc201c2e2aea7e0819479054a2c87a0bcf883..3a436eb2e5cd1a87efbbc3d46bdc4ee129c0693f 100644 --- a/test_tipc/config/DeiT/DeiT_base_patch16_224_train_infer_python.txt +++ b/test_tipc/config/DeiT/DeiT_base_patch16_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DeiT/DeiT_base_patch16_384_train_amp_infer_python.txt b/test_tipc/config/DeiT/DeiT_base_patch16_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..964e8eadbcda29764c28d646a0f1984cf85f8195 --- /dev/null +++ b/test_tipc/config/DeiT/DeiT_base_patch16_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_base_patch16_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_base_patch16_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_base_patch16_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DeiT/DeiT_base_patch16_384_train_infer_python.txt b/test_tipc/config/DeiT/DeiT_base_patch16_384_train_infer_python.txt index c8fc60e0a517c53bbd27c3cc21528eb0de61adc9..c1ee64b3bfe6482e72f10803e8fd955ae9c19896 100644 --- a/test_tipc/config/DeiT/DeiT_base_patch16_384_train_infer_python.txt +++ b/test_tipc/config/DeiT/DeiT_base_patch16_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] \ No newline at end of file diff --git a/test_tipc/config/DeiT/DeiT_small_patch16_224_train_amp_infer_python.txt b/test_tipc/config/DeiT/DeiT_small_patch16_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..32b63d774f447937d0863d7b8ffb85b290e15a2e --- /dev/null +++ b/test_tipc/config/DeiT/DeiT_small_patch16_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_small_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_small_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_small_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DeiT/DeiT_small_patch16_224_train_infer_python.txt b/test_tipc/config/DeiT/DeiT_small_patch16_224_train_infer_python.txt index 83ea4c31391ab7e07e286e38532d39e33762d171..37bdad714cd9bfead43c0f984b4cb2c36f206d4e 100644 --- a/test_tipc/config/DeiT/DeiT_small_patch16_224_train_infer_python.txt +++ b/test_tipc/config/DeiT/DeiT_small_patch16_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_amp_infer_python.txt b/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..769f086946997e78bd0f8b0bfd83512d855c6ed3 --- /dev/null +++ b/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DeiT_tiny_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DeiT/DeiT_tiny_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DeiT_tiny_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_infer_python.txt b/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_infer_python.txt index cb1850234f66e164216faa0978afcf7c95ed6d96..e262b871651de68e80e37264a334040e923521bd 100644 --- a/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_infer_python.txt +++ b/test_tipc/config/DeiT/DeiT_tiny_patch16_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DenseNet/DenseNet121_train_amp_infer_python.txt b/test_tipc/config/DenseNet/DenseNet121_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..e66857b9d6f5527efa213faf2489c45d52a7eef6 --- /dev/null +++ b/test_tipc/config/DenseNet/DenseNet121_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet121 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet121.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet121.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet121.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet121_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DenseNet/DenseNet121_train_infer_python.txt b/test_tipc/config/DenseNet/DenseNet121_train_infer_python.txt index 9f89e4cb10ef1eac5cb58910d2fba2605938f1e3..17a24fbaf250ce7d9988f15c48847efdc663606d 100644 --- a/test_tipc/config/DenseNet/DenseNet121_train_infer_python.txt +++ b/test_tipc/config/DenseNet/DenseNet121_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DenseNet/DenseNet161_train_amp_infer_python.txt b/test_tipc/config/DenseNet/DenseNet161_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..fa1a0a7e33973a575d379ec0b02b9fd6c7eef217 --- /dev/null +++ b/test_tipc/config/DenseNet/DenseNet161_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet161 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet161.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet161.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet161.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet161_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DenseNet/DenseNet161_train_infer_python.txt b/test_tipc/config/DenseNet/DenseNet161_train_infer_python.txt index 379cc29e7b9b84e56d68d34eaae0b58f3b9f8435..3d2c8519e61a5c6f5e2c819d50d53012e097f01d 100644 --- a/test_tipc/config/DenseNet/DenseNet161_train_infer_python.txt +++ b/test_tipc/config/DenseNet/DenseNet161_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DenseNet/DenseNet169_train_amp_infer_python.txt b/test_tipc/config/DenseNet/DenseNet169_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..55c4d6ec86540c175832c61af03006c5813ee184 --- /dev/null +++ b/test_tipc/config/DenseNet/DenseNet169_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet169 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet169.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet169.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet169.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet169_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DenseNet/DenseNet169_train_infer_python.txt b/test_tipc/config/DenseNet/DenseNet169_train_infer_python.txt index c2e2bbb5bd671cf717313bf393814cf146ef88b0..f962f66e4fbe71f7fcdfa355b8af101a328c83d9 100644 --- a/test_tipc/config/DenseNet/DenseNet169_train_infer_python.txt +++ b/test_tipc/config/DenseNet/DenseNet169_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DenseNet/DenseNet201_train_amp_infer_python.txt b/test_tipc/config/DenseNet/DenseNet201_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..58250ce15a9227e9adcb8c7f2faaf89ded4b7062 --- /dev/null +++ b/test_tipc/config/DenseNet/DenseNet201_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet201 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet201.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet201.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet201.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet201_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DenseNet/DenseNet201_train_infer_python.txt b/test_tipc/config/DenseNet/DenseNet201_train_infer_python.txt index 08dedc4d0ba71a45b22504990d381c2a5d0f37ca..b1bc6ad7b844826b5752ab64ee70010ce15e8d03 100644 --- a/test_tipc/config/DenseNet/DenseNet201_train_infer_python.txt +++ b/test_tipc/config/DenseNet/DenseNet201_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/DenseNet/DenseNet264_train_amp_infer_python.txt b/test_tipc/config/DenseNet/DenseNet264_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..26138926314320c8c9b5ee3831a06e036142d689 --- /dev/null +++ b/test_tipc/config/DenseNet/DenseNet264_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:DenseNet264 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/DenseNet/DenseNet264.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/DenseNet/DenseNet264.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/DenseNet/DenseNet264.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet264_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/DenseNet/DenseNet264_train_infer_python.txt b/test_tipc/config/DenseNet/DenseNet264_train_infer_python.txt index 146ad68cba3d115928fa4d09650d400f060189b5..ff95b9d8e046d098c6f36ff54596e181b3f6e934 100644 --- a/test_tipc/config/DenseNet/DenseNet264_train_infer_python.txt +++ b/test_tipc/config/DenseNet/DenseNet264_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Distillation/resnet34_distill_resnet18_dkd_train_amp_infer_python.txt b/test_tipc/config/Distillation/resnet34_distill_resnet18_dkd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ab94039471ba1abaf035600a3351656a3f4e0f25 --- /dev/null +++ b/test_tipc/config/Distillation/resnet34_distill_resnet18_dkd_train_amp_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:DistillationModel +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=100 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/Distillation/resnet34_distill_resnet18_dkd_train_infer_python.txt b/test_tipc/config/Distillation/resnet34_distill_resnet18_dkd_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4b216a9f0c2fa15811617575aefa772aa7dab313 --- /dev/null +++ b/test_tipc/config/Distillation/resnet34_distill_resnet18_dkd_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:DistillationModel +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=100 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Distillation/resnet34_distill_resnet18_dkd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/ESNet/ESNet_x0_25_train_amp_infer_python.txt b/test_tipc/config/ESNet/ESNet_x0_25_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..7e9abcf47db884882e23008c9909dd3f334f4773 --- /dev/null +++ b/test_tipc/config/ESNet/ESNet_x0_25_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ESNet/ESNet_x0_25_train_infer_python.txt b/test_tipc/config/ESNet/ESNet_x0_25_train_infer_python.txt index 82ead10440971c0fad9b44606fd957a0fb20c984..df1f896da971a61dd3743830a3b25dfd50b087d2 100644 --- a/test_tipc/config/ESNet/ESNet_x0_25_train_infer_python.txt +++ b/test_tipc/config/ESNet/ESNet_x0_25_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ESNet/ESNet_x0_5_train_amp_infer_python.txt b/test_tipc/config/ESNet/ESNet_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..7cd576052d101c9dbc065cfe6dae6a290a5f1490 --- /dev/null +++ b/test_tipc/config/ESNet/ESNet_x0_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ESNet/ESNet_x0_5_train_infer_python.txt b/test_tipc/config/ESNet/ESNet_x0_5_train_infer_python.txt index 38b376f3052020a8c810497f23b640c4167f37e0..b8aa0bc75ed239c7f106607f9f67b4fe6891035e 100644 --- a/test_tipc/config/ESNet/ESNet_x0_5_train_infer_python.txt +++ b/test_tipc/config/ESNet/ESNet_x0_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ESNet/ESNet_x0_75_train_amp_infer_python.txt b/test_tipc/config/ESNet/ESNet_x0_75_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..aa715a5d0165b93e5d4808353ac4a75cda974064 --- /dev/null +++ b/test_tipc/config/ESNet/ESNet_x0_75_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ESNet/ESNet_x0_75_train_infer_python.txt b/test_tipc/config/ESNet/ESNet_x0_75_train_infer_python.txt index dc6349809b771211e323083cfc7638697086cfec..713086fe05ea65d33d53f700d363349d96e365c7 100644 --- a/test_tipc/config/ESNet/ESNet_x0_75_train_infer_python.txt +++ b/test_tipc/config/ESNet/ESNet_x0_75_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ESNet/ESNet_x1_0_train_amp_infer_python.txt b/test_tipc/config/ESNet/ESNet_x1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..8ce5c4ca603eca59b8f4d7df3deb2fa31ca0ad3f --- /dev/null +++ b/test_tipc/config/ESNet/ESNet_x1_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ESNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ESNet/ESNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ESNet/ESNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ESNet/ESNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ESNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ESNet/ESNet_x1_0_train_infer_python.txt b/test_tipc/config/ESNet/ESNet_x1_0_train_infer_python.txt index c5a4a5af55b18c398776f64595c3e89349cce111..31e11cff36f60eea65c97e7431b6f39a60d9695e 100644 --- a/test_tipc/config/ESNet/ESNet_x1_0_train_infer_python.txt +++ b/test_tipc/config/ESNet/ESNet_x1_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB0_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..effea01493bf4ee309feea2ab71947117621df74 --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB0_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB0_train_infer_python.txt index 23b3ef1c06627a374fb8c585906942f4bf873744..fe9660749f84853ac5e9111427ca43cfe8740bac 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB0_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB1_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB1_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ba5fa4d2067ddd98058ecc0b762292bf27219cb7 --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB1_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB1.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB1_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=272 -o PreProcess.transform_ops.1.CropImage.size=240 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB1_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB1_train_infer_python.txt index 1c4314548c4bf647413efdec3e791caa16b21f32..302b52b5175a5f012c4d0c4c0ca993e0a26baca0 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB1_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB1_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,240,240]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB2_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB2_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a045ea9fc1ff69b039b78f92dd448e3fc53927cb --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB2_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB2.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB2_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=292 -o PreProcess.transform_ops.1.CropImage.size=260 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB2_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB2_train_infer_python.txt index 0ac16c3d1020f2b8cf719de0aa19a5151f5c1a0a..116f9ca21a6294815a6dd1dd57406e43fc029e95 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB2_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB2_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,260,260]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB3_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB3_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..dc3a03297aee923f0476c6193b14482da89292bb --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB3_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=332 -o PreProcess.transform_ops.1.CropImage.size=300 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB3_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB3_train_infer_python.txt index 0d0ca3dbf4cf8bda606cbad688a9d225e3159bdc..3e5369b56b3ba2d0707120b1f07e71262cf49cc5 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB3_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB3_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,300,300]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB4_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB4_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1c63f3b742101adff3015614782065a3a6f64094 --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB4_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB4 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB4.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB4.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB4.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB4_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=412 -o PreProcess.transform_ops.1.CropImage.size=380 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB4_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB4_train_infer_python.txt index 92d4cb818a26b9c8d1a1c88af438780ebd3b49ba..51b439e459d57800904b6aadc7f7c44ec37038c4 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB4_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB4_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,380,380]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB5_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..60d0eeac26259f0a986332dfb19b849a25d2ae08 --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=488 -o PreProcess.transform_ops.1.CropImage.size=456 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB5_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB5_train_infer_python.txt index 02c935746c060dfd0126d002bafa05e9e5d6a3fd..77ac03863e989f359b1f3f5fa063da4ed721a4f1 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB5_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,456,456]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB6_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB6_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ce07cce099b56e668e54672a4783ebaa8041fe25 --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB6_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB6 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB6.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB6.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB6.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB6_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=560 -o PreProcess.transform_ops.1.CropImage.size=528 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB6_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB6_train_infer_python.txt index 392c0482bdcc70846613ab0f0f425b3290ae81df..73c88c4dfbb975dd5585e207b6d9c09017de94b2 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB6_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB6_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,528,528]}] \ No newline at end of file diff --git a/test_tipc/config/EfficientNet/EfficientNetB7_train_amp_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB7_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..d8eff7add7411b105b0996c5d91166e4689f6ef4 --- /dev/null +++ b/test_tipc/config/EfficientNet/EfficientNetB7_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:EfficientNetB7 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB7.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB7.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/EfficientNet/EfficientNetB7.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/EfficientNetB7_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=632 -o PreProcess.transform_ops.1.CropImage.size=600 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/EfficientNet/EfficientNetB7_train_infer_python.txt b/test_tipc/config/EfficientNet/EfficientNetB7_train_infer_python.txt index 7265243e6d61a603d56f311159bb98e203256e9e..dd70d66b2a0a48ce1f33104bbcfef383b82a3258 100644 --- a/test_tipc/config/EfficientNet/EfficientNetB7_train_infer_python.txt +++ b/test_tipc/config/EfficientNet/EfficientNetB7_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tra -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,600,600]}] \ No newline at end of file diff --git a/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_amp_infer_python.txt b/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..2420e06e03fa4abe66e678a7a06940a391a590a9 --- /dev/null +++ b/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GeneralRecognition_PPLCNet_x2_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/rec/models/pretrain/general_PPLCNet_x2_5_pretrained_v1.0.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_rec.py -c configs/inference_rec.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.rec_inference_model_dir:../inference +-o Global.infer_imgs:../dataset/Aliproduct/demo_test/ +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_infer_python.txt b/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_infer_python.txt index 40e6675a6717ad1448ea0663f0152fbefd5987c7..cdcb9f737d8c1f6be29ff38f9be187d6130a1c7b 100644 --- a/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_infer_python.txt +++ b/test_tipc/config/GeneralRecognition/GeneralRecognition_PPLCNet_x2_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_rec.py -c configs/inference_rec.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/GhostNet/GhostNet_x0_5_train_amp_infer_python.txt b/test_tipc/config/GhostNet/GhostNet_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0861f243efb67150067ff694cc7be2fe11ef080c --- /dev/null +++ b/test_tipc/config/GhostNet/GhostNet_x0_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GhostNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/GhostNet/GhostNet_x0_5_train_infer_python.txt b/test_tipc/config/GhostNet/GhostNet_x0_5_train_infer_python.txt index 7df8dee4dd9b66ba9c4bdb51f7b05f512f86f0b2..5207cdbee52e14412523e6124d94b070b41f8036 100644 --- a/test_tipc/config/GhostNet/GhostNet_x0_5_train_infer_python.txt +++ b/test_tipc/config/GhostNet/GhostNet_x0_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/GhostNet/GhostNet_x1_0_train_amp_infer_python.txt b/test_tipc/config/GhostNet/GhostNet_x1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3132830ca27910adce290b231257ebd4e1506f79 --- /dev/null +++ b/test_tipc/config/GhostNet/GhostNet_x1_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GhostNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/GhostNet/GhostNet_x1_0_train_infer_python.txt b/test_tipc/config/GhostNet/GhostNet_x1_0_train_infer_python.txt index 7d3184b4f06bca8eae2cff9523b2b3e6991e5c79..600f238902823521de8ab7816d4fab92d25be9c5 100644 --- a/test_tipc/config/GhostNet/GhostNet_x1_0_train_infer_python.txt +++ b/test_tipc/config/GhostNet/GhostNet_x1_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/GhostNet/GhostNet_x1_3_train_amp_infer_python.txt b/test_tipc/config/GhostNet/GhostNet_x1_3_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..e5ad93bc1fc381add7241943744f731f6cb75bf4 --- /dev/null +++ b/test_tipc/config/GhostNet/GhostNet_x1_3_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GhostNet_x1_3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/GhostNet/GhostNet_x1_3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GhostNet_x1_3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/GhostNet/GhostNet_x1_3_train_infer_python.txt b/test_tipc/config/GhostNet/GhostNet_x1_3_train_infer_python.txt index e186eea4c126bb38d5272aa5c650818a10c96a90..f859260b60b4d830ff1c88b468e736b3074fc945 100644 --- a/test_tipc/config/GhostNet/GhostNet_x1_3_train_infer_python.txt +++ b/test_tipc/config/GhostNet/GhostNet_x1_3_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HRNet/HRNet_W18_C_train_amp_infer_python.txt b/test_tipc/config/HRNet/HRNet_W18_C_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..72014b2741530ca16aa508bc61e1f6d6253f0d20 --- /dev/null +++ b/test_tipc/config/HRNet/HRNet_W18_C_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W18_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W18_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W18_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HRNet/HRNet_W18_C_train_infer_python.txt b/test_tipc/config/HRNet/HRNet_W18_C_train_infer_python.txt index 3883580f01290343ff95b0caa23d11b8b8edef47..a98161cd26ad75b589778c79144428c946066703 100644 --- a/test_tipc/config/HRNet/HRNet_W18_C_train_infer_python.txt +++ b/test_tipc/config/HRNet/HRNet_W18_C_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HRNet/HRNet_W30_C_train_amp_infer_python.txt b/test_tipc/config/HRNet/HRNet_W30_C_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0a38ecc036167f4d35592d78ae9b51c63e096cce --- /dev/null +++ b/test_tipc/config/HRNet/HRNet_W30_C_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W30_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W30_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W30_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W30_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W30_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HRNet/HRNet_W30_C_train_infer_python.txt b/test_tipc/config/HRNet/HRNet_W30_C_train_infer_python.txt index cba0265648d5756645e48c2059bafcbe31ac7b04..6376de04fd54dd337d018e26b58da0e8a67302a2 100644 --- a/test_tipc/config/HRNet/HRNet_W30_C_train_infer_python.txt +++ b/test_tipc/config/HRNet/HRNet_W30_C_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HRNet/HRNet_W32_C_train_amp_infer_python.txt b/test_tipc/config/HRNet/HRNet_W32_C_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..bcdd374618799e8f7cc519f77f40b36ad3a57e51 --- /dev/null +++ b/test_tipc/config/HRNet/HRNet_W32_C_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W32_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W32_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W32_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W32_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W32_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HRNet/HRNet_W32_C_train_infer_python.txt b/test_tipc/config/HRNet/HRNet_W32_C_train_infer_python.txt index 197a76bab6d08e31260e1daee36a218370198240..e4aef325e33d326fbbffdcfe5ad5ea2d072f0e62 100644 --- a/test_tipc/config/HRNet/HRNet_W32_C_train_infer_python.txt +++ b/test_tipc/config/HRNet/HRNet_W32_C_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HRNet/HRNet_W40_C_train_amp_infer_python.txt b/test_tipc/config/HRNet/HRNet_W40_C_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3825c10286046c98b06aadc17d1581e293967d59 --- /dev/null +++ b/test_tipc/config/HRNet/HRNet_W40_C_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W40_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W40_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W40_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W40_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W40_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HRNet/HRNet_W40_C_train_infer_python.txt b/test_tipc/config/HRNet/HRNet_W40_C_train_infer_python.txt index 1deb8917c7c00e07a36f237392494687b215a2c1..9ad57d9b0be126b3e8bf538a7c9dfd43b59fe109 100644 --- a/test_tipc/config/HRNet/HRNet_W40_C_train_infer_python.txt +++ b/test_tipc/config/HRNet/HRNet_W40_C_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HRNet/HRNet_W44_C_train_amp_infer_python.txt b/test_tipc/config/HRNet/HRNet_W44_C_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f7bdb981a921b7971969b25e28c2ab56535546f6 --- /dev/null +++ b/test_tipc/config/HRNet/HRNet_W44_C_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W44_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W44_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W44_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W44_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W44_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HRNet/HRNet_W44_C_train_infer_python.txt b/test_tipc/config/HRNet/HRNet_W44_C_train_infer_python.txt index 10fab3d371c57305c576b514b83e83959d7e98f3..14710ed597eb732b1ef9e8ebb5a0bcaf3ba8577b 100644 --- a/test_tipc/config/HRNet/HRNet_W44_C_train_infer_python.txt +++ b/test_tipc/config/HRNet/HRNet_W44_C_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HRNet/HRNet_W48_C_train_amp_infer_python.txt b/test_tipc/config/HRNet/HRNet_W48_C_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..56cc623d73b357c873d3f75535cfc62882030db5 --- /dev/null +++ b/test_tipc/config/HRNet/HRNet_W48_C_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W48_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W48_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W48_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W48_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W48_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HRNet/HRNet_W48_C_train_infer_python.txt b/test_tipc/config/HRNet/HRNet_W48_C_train_infer_python.txt index 6de8e26ae125309632a04cf3ae2a557464254017..f85ce5a5e2893c1b4c41e2370802dc3d36a55bd7 100644 --- a/test_tipc/config/HRNet/HRNet_W48_C_train_infer_python.txt +++ b/test_tipc/config/HRNet/HRNet_W48_C_train_infer_python.txt @@ -17,7 +17,7 @@ norm_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W48_C.yaml -o Gl pact_train:null fpgm_train:null distill_train:null -null:null +to_static_train:-o Global.to_static=True null:null ## ===========================eval_params=========================== @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:64|128 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HRNet/HRNet_W64_C_train_amp_infer_python.txt b/test_tipc/config/HRNet/HRNet_W64_C_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..daf5d0ab12fc1c0fd3a29a677dd178e5a16c1432 --- /dev/null +++ b/test_tipc/config/HRNet/HRNet_W64_C_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HRNet_W64_C +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HRNet/HRNet_W64_C.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HRNet/HRNet_W64_C.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HRNet/HRNet_W64_C.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/HRNet_W64_C_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HRNet/HRNet_W64_C_train_infer_python.txt b/test_tipc/config/HRNet/HRNet_W64_C_train_infer_python.txt index c99c5e6c00cad8b5a2e867b58da13b8016f91552..75b19de8f3bd36fe21db217a211d066f3d8516aa 100644 --- a/test_tipc/config/HRNet/HRNet_W64_C_train_infer_python.txt +++ b/test_tipc/config/HRNet/HRNet_W64_C_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HarDNet/HarDNet39_ds_train_amp_infer_python.txt b/test_tipc/config/HarDNet/HarDNet39_ds_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..122dafe1a411b40210ae74864aee4e1e2b1f0f3e --- /dev/null +++ b/test_tipc/config/HarDNet/HarDNet39_ds_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet39_ds +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet39_ds.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet39_ds.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet39_ds.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet39_ds_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HarDNet/HarDNet39_ds_train_infer_python.txt b/test_tipc/config/HarDNet/HarDNet39_ds_train_infer_python.txt index 5cda671bfcae9129facd62a93c2657d87f192d62..1d2b13eccfbbbca76550115748c5213dac8b32e4 100644 --- a/test_tipc/config/HarDNet/HarDNet39_ds_train_infer_python.txt +++ b/test_tipc/config/HarDNet/HarDNet39_ds_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HarDNet/HarDNet68_ds_train_amp_infer_python.txt b/test_tipc/config/HarDNet/HarDNet68_ds_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..8d27c482d6496430a0f982e0031cb503036caab4 --- /dev/null +++ b/test_tipc/config/HarDNet/HarDNet68_ds_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet68_ds +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68_ds.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68_ds.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68_ds.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_ds_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HarDNet/HarDNet68_ds_train_infer_python.txt b/test_tipc/config/HarDNet/HarDNet68_ds_train_infer_python.txt index bfe601ee0a0d953ca8f20b9d46c5a146c98fdbe4..0e6172a0cd768060e15bd06052a2b9fe22a5ea8e 100644 --- a/test_tipc/config/HarDNet/HarDNet68_ds_train_infer_python.txt +++ b/test_tipc/config/HarDNet/HarDNet68_ds_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HarDNet/HarDNet68_train_amp_infer_python.txt b/test_tipc/config/HarDNet/HarDNet68_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..88c82cc41e2c09ee53ca6999a8c14a64125db6e3 --- /dev/null +++ b/test_tipc/config/HarDNet/HarDNet68_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet68 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet68.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet68_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HarDNet/HarDNet68_train_infer_python.txt b/test_tipc/config/HarDNet/HarDNet68_train_infer_python.txt index b61ea632e23f51af6a0bb15fb8c6ac3af91caaf4..7211e8b738aaf5c636606528a98dcee404ee1056 100644 --- a/test_tipc/config/HarDNet/HarDNet68_train_infer_python.txt +++ b/test_tipc/config/HarDNet/HarDNet68_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/HarDNet/HarDNet85_train_amp_infer_python.txt b/test_tipc/config/HarDNet/HarDNet85_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..5c3a73dfa2eacb6b35c2dd296dc3126d921acd75 --- /dev/null +++ b/test_tipc/config/HarDNet/HarDNet85_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:HarDNet85 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/HarDNet/HarDNet85.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/HarDNet/HarDNet85.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/HarDNet/HarDNet85.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/HarDNet85_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/HarDNet/HarDNet85_train_infer_python.txt b/test_tipc/config/HarDNet/HarDNet85_train_infer_python.txt index 02237ba7e2584c494821633d94e4302fc7bf0dc5..5371fe9321135a7c953e81a3d9f42d1c456c86b4 100644 --- a/test_tipc/config/HarDNet/HarDNet85_train_infer_python.txt +++ b/test_tipc/config/HarDNet/HarDNet85_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Inception/GoogLeNet_train_amp_infer_python.txt b/test_tipc/config/Inception/GoogLeNet_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..de01f30b696586f484f7b1921536a3189eb588ae --- /dev/null +++ b/test_tipc/config/Inception/GoogLeNet_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:GoogLeNet +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Inception/GoogLeNet.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Inception/GoogLeNet.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Inception/GoogLeNet.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/GoogLeNet_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Inception/GoogLeNet_train_infer_python.txt b/test_tipc/config/Inception/GoogLeNet_train_infer_python.txt index ebbb86dd714aeca81a051c13e60828f516c50577..0ce44c563ea95ddf41630394d39fd8f42e7a6924 100644 --- a/test_tipc/config/Inception/GoogLeNet_train_infer_python.txt +++ b/test_tipc/config/Inception/GoogLeNet_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Inception/InceptionV3_train_amp_infer_python.txt b/test_tipc/config/Inception/InceptionV3_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ce9c3477bef4287532d6169b6e1f8780160d38ae --- /dev/null +++ b/test_tipc/config/Inception/InceptionV3_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:InceptionV3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Inception/InceptionV3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Inception/InceptionV3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Inception/InceptionV3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Inception/InceptionV3_train_infer_python.txt b/test_tipc/config/Inception/InceptionV3_train_infer_python.txt index 6a7960c671cafae7c63268366c0e0309b90f4cd8..64e857a5878ca79763e749562a988325ecbf5569 100644 --- a/test_tipc/config/Inception/InceptionV3_train_infer_python.txt +++ b/test_tipc/config/Inception/InceptionV3_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tra -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,299,299]}] \ No newline at end of file diff --git a/test_tipc/config/Inception/InceptionV4_train_amp_infer_python.txt b/test_tipc/config/Inception/InceptionV4_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..853c1b7da138ecdbea17492e2e5d83515f9887fc --- /dev/null +++ b/test_tipc/config/Inception/InceptionV4_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:InceptionV4 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Inception/InceptionV4.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Inception/InceptionV4.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Inception/InceptionV4.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/InceptionV4_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Inception/InceptionV4_train_infer_python.txt b/test_tipc/config/Inception/InceptionV4_train_infer_python.txt index 1e1d7fd5ddee6fffa34f6501d7b10f82ef2b707d..75826301ef65fbe0fc4dcf3fe2407deb4881e099 100644 --- a/test_tipc/config/Inception/InceptionV4_train_infer_python.txt +++ b/test_tipc/config/Inception/InceptionV4_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tra -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,299,299]}] \ No newline at end of file diff --git a/test_tipc/config/LeViT/LeViT_128S_train_amp_infer_python.txt b/test_tipc/config/LeViT/LeViT_128S_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..d973ec5174940d92c629bf5be745a74c4d73a3d7 --- /dev/null +++ b/test_tipc/config/LeViT/LeViT_128S_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_128S +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_128S.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128S_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/LeViT/LeViT_128S_train_infer_python.txt b/test_tipc/config/LeViT/LeViT_128S_train_infer_python.txt index 772049b8870e8413407522b38b0781279eb2f740..0b3daa1e6f05c8dc086cd09397f0311054f61fee 100644 --- a/test_tipc/config/LeViT/LeViT_128S_train_infer_python.txt +++ b/test_tipc/config/LeViT/LeViT_128S_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/LeViT/LeViT_128_train_amp_infer_python.txt b/test_tipc/config/LeViT/LeViT_128_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..fb884b24b97cec70913380aee85f28e0c1d66f5f --- /dev/null +++ b/test_tipc/config/LeViT/LeViT_128_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_128 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_128.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_128.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_128.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_128_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/LeViT/LeViT_128_train_infer_python.txt b/test_tipc/config/LeViT/LeViT_128_train_infer_python.txt index 8c8699edc09206a8d6bda35e01bcfeeef17cac56..515b109d74b2e4000dd58d57945cd4e2a4024955 100644 --- a/test_tipc/config/LeViT/LeViT_128_train_infer_python.txt +++ b/test_tipc/config/LeViT/LeViT_128_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/LeViT/LeViT_192_train_amp_infer_python.txt b/test_tipc/config/LeViT/LeViT_192_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0ecb7e096a8983b714d489a15176df92ebd2b1fa --- /dev/null +++ b/test_tipc/config/LeViT/LeViT_192_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_192 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_192.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_192.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_192.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_192_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/LeViT/LeViT_192_train_infer_python.txt b/test_tipc/config/LeViT/LeViT_192_train_infer_python.txt index c2ac115fcc4373b8db65e2d0897f7e8f034ddebb..107977ceadc93a3b40d25b01cbe311871b24675d 100644 --- a/test_tipc/config/LeViT/LeViT_192_train_infer_python.txt +++ b/test_tipc/config/LeViT/LeViT_192_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/LeViT/LeViT_256_train_amp_infer_python.txt b/test_tipc/config/LeViT/LeViT_256_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..dbea83a768e91287e98bfdf6125bf1d786f0ded5 --- /dev/null +++ b/test_tipc/config/LeViT/LeViT_256_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_256 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_256.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_256.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_256.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_256_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/LeViT/LeViT_256_train_infer_python.txt b/test_tipc/config/LeViT/LeViT_256_train_infer_python.txt index f58880c80cde3b5b13eca82b19871bbfa731bdc3..18bee51e16a6e86aa09a994abece11cbb462cf00 100644 --- a/test_tipc/config/LeViT/LeViT_256_train_infer_python.txt +++ b/test_tipc/config/LeViT/LeViT_256_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/LeViT/LeViT_384_train_amp_infer_python.txt b/test_tipc/config/LeViT/LeViT_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..c7a243e79f07e62cc3c39680398889101bd2cb1d --- /dev/null +++ b/test_tipc/config/LeViT/LeViT_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:LeViT_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/LeViT/LeViT_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/LeViT/LeViT_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/LeViT/LeViT_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/LeViT_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=256 -o PreProcess.transform_ops.1.CropImage.size=224 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/LeViT/LeViT_384_train_infer_python.txt b/test_tipc/config/LeViT/LeViT_384_train_infer_python.txt index 0ae0d40fbfcd9a9a1c8ea187c534943613228d75..968b767dbd504917ae4ef93aad914bffe8e95606 100644 --- a/test_tipc/config/LeViT/LeViT_384_train_infer_python.txt +++ b/test_tipc/config/LeViT/LeViT_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MixNet/MixNet_L_train_amp_infer_python.txt b/test_tipc/config/MixNet/MixNet_L_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a70575726f4690e0abe8c698c7674918a3c64dc4 --- /dev/null +++ b/test_tipc/config/MixNet/MixNet_L_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MixNet_L +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MixNet/MixNet_L.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MixNet/MixNet_L.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MixNet/MixNet_L.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_L_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MixNet/MixNet_L_train_infer_python.txt b/test_tipc/config/MixNet/MixNet_L_train_infer_python.txt index 46fcbc1dc38eacb7d109fb0d157569a224593dae..26e2ba0ce3f32c932196a7c338b00da29f9b8265 100644 --- a/test_tipc/config/MixNet/MixNet_L_train_infer_python.txt +++ b/test_tipc/config/MixNet/MixNet_L_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MixNet/MixNet_M_train_amp_infer_python.txt b/test_tipc/config/MixNet/MixNet_M_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..66c5e83d7d6a64281e3c8c48f3a75641dd2ba826 --- /dev/null +++ b/test_tipc/config/MixNet/MixNet_M_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MixNet_M +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MixNet/MixNet_M.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MixNet/MixNet_M.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MixNet/MixNet_M.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_M_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MixNet/MixNet_M_train_infer_python.txt b/test_tipc/config/MixNet/MixNet_M_train_infer_python.txt index 87be78a543227d36bb90e11e1151dd0626647b96..c214185178d2ce17df7068cf53a73948de9f42df 100644 --- a/test_tipc/config/MixNet/MixNet_M_train_infer_python.txt +++ b/test_tipc/config/MixNet/MixNet_M_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MixNet/MixNet_S_train_amp_infer_python.txt b/test_tipc/config/MixNet/MixNet_S_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ac256e161cf75a7dc1d3978af7ba2a64fd1c2cc9 --- /dev/null +++ b/test_tipc/config/MixNet/MixNet_S_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MixNet_S +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MixNet/MixNet_S.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MixNet/MixNet_S.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MixNet/MixNet_S.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MixNet_S_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MixNet/MixNet_S_train_infer_python.txt b/test_tipc/config/MixNet/MixNet_S_train_infer_python.txt index 42e5df7f087474d55cb2e8e19163fd054537f43b..77cb00769b7c15b4ef3473b63a19e584490dea3c 100644 --- a/test_tipc/config/MixNet/MixNet_S_train_infer_python.txt +++ b/test_tipc/config/MixNet/MixNet_S_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_train_amp_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..db98a5f6ed047a7236d0a7a1bbfa91e0f5093995 --- /dev/null +++ b/test_tipc/config/MobileNetV1/MobileNetV1_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +to_static_train:-o Global.to_static=True +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_train_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_train_infer_python.txt index 65fee1526d93b04ff634e44879dc9449cd9df607..ac12953ddb4070265b7ebe714182c79fe070138a 100644 --- a/test_tipc/config/MobileNetV1/MobileNetV1_train_infer_python.txt +++ b/test_tipc/config/MobileNetV1/MobileNetV1_train_infer_python.txt @@ -17,7 +17,7 @@ norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1.yaml pact_train:null fpgm_train:null distill_train:null -null:null +to_static_train:-o Global.to_static=True null:null ## ===========================eval_params=========================== @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:64|128 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_amp_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..68106952dfb6612916963c5c1f6d7c932eeffe23 --- /dev/null +++ b/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_infer_python.txt index a9f63dfab3c39fe13cf7f0aebf8fd68a70011a21..5dd3399159d6377cab5e38ce101e1f128cac0ca2 100644 --- a/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_infer_python.txt +++ b/test_tipc/config/MobileNetV1/MobileNetV1_x0_25_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_amp_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0552171764cdaeb448267cf5c95988354d0691bc --- /dev/null +++ b/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_infer_python.txt index 379cbd665c3f2df0d7adc5804d043be411eeb7c3..fa8e48f42357930f214c23190f947d5de2df5f61 100644 --- a/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_infer_python.txt +++ b/test_tipc/config/MobileNetV1/MobileNetV1_x0_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_amp_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4d876f8fb50e93fdfd53a9bcbb2f8c8f13a687e4 --- /dev/null +++ b/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV1_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV1/MobileNetV1_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV1_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_infer_python.txt b/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_infer_python.txt index fef81fa3e819957572eeb511a30b579b90b1f1bd..f775c3f850ff35043920f478ecb66b5f5c77d31e 100644 --- a/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_infer_python.txt +++ b/test_tipc/config/MobileNetV1/MobileNetV1_x0_75_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_train_amp_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..840a0f598abf0bd047f7ed8cf7085974fb856e5f --- /dev/null +++ b/test_tipc/config/MobileNetV2/MobileNetV2_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +to_static_train:-o Global.to_static=True +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt index fb0c8a24716bdcb7506e3d3e67a31fedeb2929f4..eef9b575203102f55d335c43ba6bda353581df53 100644 --- a/test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt +++ b/test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt @@ -17,7 +17,7 @@ norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2.yaml pact_train:null fpgm_train:null distill_train:null -null:null +to_static_train:-o Global.to_static=True null:null ## ===========================eval_params=========================== @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:64|128 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_amp_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1c2718f405a059c9eebdd7e0ca60edd17e065a8a --- /dev/null +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_infer_python.txt index 9183e01e10ae7f63c2e01bfa3eaf5c2d0cc6fe4f..98830b6ed6569e541c1d8c35d69da9c1370cefd4 100644 --- a/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_infer_python.txt +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x0_25_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_amp_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..03f8073e5e1047be5024dc04c7a34ba8943246c0 --- /dev/null +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_infer_python.txt index bef4e1d924e93b22d602236723f2fdf85ab177fa..8593560ffba05496fccee465854aac5c9ab158d3 100644 --- a/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_infer_python.txt +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x0_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_amp_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..93547f8dc4e19ffa13bb4662619e9fb908a67dbb --- /dev/null +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_infer_python.txt index a788731671f47f46ae59658ddf90d0af58597e4d..657e76a3f82605bf0f443078b8bb15a8e9ec182b 100644 --- a/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_infer_python.txt +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x0_75_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_amp_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..948db736c5d1e287a1d8c9378f1106365a30a3ad --- /dev/null +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_infer_python.txt index d5f9bf4113127db437e535a488316597c9ccf00e..f49d684ad70cb88b36df6e2c7c095a26e97b329c 100644 --- a/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_infer_python.txt +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x1_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_amp_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..6b1b20096b72760432ee59dc6ebc21028eee1b86 --- /dev/null +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV2_x2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV2/MobileNetV2_x2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV2_x2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_infer_python.txt b/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_infer_python.txt index f445570483c9d802ca70725364dbc9021be3bbf8..7fc9c6e9108f32f310e6ebff12c6191205164476 100644 --- a/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_infer_python.txt +++ b/test_tipc/config/MobileNetV2/MobileNetV2_x2_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..72c0b5f4de86fd39d1c634f8d7f1afee36ec74ff --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_infer_python.txt index 2797a2d249071e945b9ff17515622262a7f2a5dc..138b0b82245d24cd2ed19c2afd38751458c0bf11 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_35_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..28cd39dd69bfbb065cbfae257892ad9540d5a5e3 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_infer_python.txt index 6f9165ca93834464ceb4597c565c7eb6178acd21..a01e1b1ca252fa2309522563d0f65f9002c49a0f 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_5_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f3749084578921829b31a896adc2906d0afc2ca4 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_infer_python.txt index da486711c39b42b808bb7b964fe947a0333e59a2..90dccf305200619b4b573b4e88df7720bc58b39a 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x0_75_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_FPGM_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_FPGM_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..5c6dd683c4dfde02de4747e9429bdb6a9f37a1f2 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_FPGM_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_0_FPGM +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_PACT_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_PACT_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..20f2d7b947f54744be0789742b5196495f7816e0 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_PACT_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_0_PACT +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_lite_arm_cpu_cpp.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..b45c2a01b2cb0d9491b516e2caf410ef04e7d35e --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:MobileNetV3_large_x1_0 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..61e864c79e0b905e55d0c83e90980f90e1ae28db --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_0.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt index 25ecbd5f443c4077096d882876b1cae50cf47042..980f6226df6998d2cbc0a0858f628965525f41f9 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt @@ -17,7 +17,7 @@ norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_larg pact_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False fpgm_train:tools/train.py -c ppcls/configs/slim/MobileNetV3_large_x1_0_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False distill_train:null -null:null +to_static_train:-o Global.to_static=True null:null ## ===========================eval_params=========================== @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:256|640 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3b3ee3adad01e951615f10d57ce1e7e43cd64315 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_large_x1_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_large_x1_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_large_x1_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_infer_python.txt index 4b62f541d1892c79f90d1ecddf3f27484d90180a..1835f2d92cbf01d4a630a0367d528687477631ba 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_large_x1_25_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b42def4c8408d4570f03ddfe3337f704227d5fe4 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_infer_python.txt index 2e40f8992b0f715f1776fcf1d4a37541731592a9..2f184e9272ace1a72186f14ab516225d68ef35a0 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_35_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1382431826b34703e0e59fd1524390a1a9b4aeb3 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_infer_python.txt index e8a2b0e1c21f4aad8403882abf236a0813e5ef1a..4289fa00d29a5b46f5044fcbbdb2dd61c350379a 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_5_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..489b9408fdf2df35159e4e349b21f8c307a5a7b7 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_infer_python.txt index 65a8671372d1bf0819ecc96d8c355c52b6a69cd3..c49efec3e8187344e4e41bbe4d0398fc2b4c8ac2 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x0_75_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..baa1546daa565741884f93121739b361dff4ad88 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_infer_python.txt index 6d2334b4804d3dda361081cc929baf22cad23c83..6741d22b3536512471aa54b0ede5a2259f2caaa7 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_0_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_amp_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..e34e17696b728c9071cf3bd6d65e4cd404b7c348 --- /dev/null +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:MobileNetV3_small_x1_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileNetV3/MobileNetV3_small_x1_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/MobileNetV3_small_x1_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_infer_python.txt b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_infer_python.txt index 6a730b7e1fb5ff879072c04498b177053c74c9da..fb6a02d2a7bd80229a1e70b8702eeb7bf3303240 100644 --- a/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_infer_python.txt +++ b/test_tipc/config/MobileNetV3/MobileNetV3_small_x1_25_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/MobileViT/MobileViT_S_train_infer_python.txt b/test_tipc/config/MobileViT/MobileViT_S_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..619416b3a86dc029f66dc0a3b47b322ca827fb90 --- /dev/null +++ b/test_tipc/config/MobileViT/MobileViT_S_train_infer_python.txt @@ -0,0 +1,60 @@ +===========================train_params=========================== +model_name:MobileViT_S +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_S.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o Global.print_batch_step=1 +pact_train:null +fpgm_train:null +distill_train:null +to_static_train:-o Global.to_static=True +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_S.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_S.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_S_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=292 -o PreProcess.transform_ops.1.CropImage.size=256 -o PreProcess.transform_ops.2.NormalizeImage.mean=[0.,0.,0.] -o PreProcess.transform_ops.2.NormalizeImage.std=[1.,1.,1.] +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================train_benchmark_params========================== +batch_size:128 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,256,256]}] diff --git a/test_tipc/config/MobileViT/MobileViT_XS_train_infer_python.txt b/test_tipc/config/MobileViT/MobileViT_XS_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b7436adaaac2a94df9692845b92bc5c46887ba62 --- /dev/null +++ b/test_tipc/config/MobileViT/MobileViT_XS_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:MobileViT_XS +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_XS.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_XS.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_XS.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XS_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=292 -o PreProcess.transform_ops.1.CropImage.size=256 -o PreProcess.transform_ops.2.NormalizeImage.mean=[0.,0.,0.] -o PreProcess.transform_ops.2.NormalizeImage.std=[1.,1.,1.] +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,256,256]}] diff --git a/test_tipc/config/MobileViT/MobileViT_XXS_train_infer_python.txt b/test_tipc/config/MobileViT/MobileViT_XXS_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..55cb442148ece745be4e2de6d95e2bcaf5fb0074 --- /dev/null +++ b/test_tipc/config/MobileViT/MobileViT_XXS_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:MobileViT_XXS +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_XXS.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_XXS.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/MobileViT/MobileViT_XXS.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileViT_XXS_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=292 -o PreProcess.transform_ops.1.CropImage.size=256 -o PreProcess.transform_ops.2.NormalizeImage.mean=[0.,0.,0.] -o PreProcess.transform_ops.2.NormalizeImage.std=[1.,1.,1.] +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,256,256]}] diff --git a/test_tipc/config/PPLCNet/MobileNetV3_large_x1_0_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/MobileNetV3_large_x1_0_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..b45c2a01b2cb0d9491b516e2caf410ef04e7d35e --- /dev/null +++ b/test_tipc/config/PPLCNet/MobileNetV3_large_x1_0_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:MobileNetV3_large_x1_0 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_25_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_25_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..e1b9fbabec8ea44d03f8fbb7819849b7d02e4697 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_25_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:PPLCNet_x0_25 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1977e30f5f91e1a39b805e205279d59e8e81e570 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_infer_python.txt index 66afc24db947928ef5ed9267bee9c85cd543d808..b1596ee33ccade9fe3e8347ded973c3ccad35b67 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_25_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ad2ceb210308db6679d0904bb3ed1c7c494839af --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_35 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_35.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_35_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_infer_python.txt index 9aa5362f2004e080dad77ae8e66e9a9db564f3e4..57d8aa324c5175380053fa3ddec12f0e867dfb99 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_35_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_5_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_5_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..8c23191a37bca25fff5d0b6919b172d580ceecc5 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_5_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:PPLCNet_x0_5 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..6f5318b8a965543cc88b4864b9e486250b1d6999 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_infer_python.txt index 277ef813387cf3067878c84624118d6e3b1cb1d9..025fa994ebab8b9e9472ef62ef8fe7edc4c57c87 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_5_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_75_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_75_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..3b1a799109836fbd6188af842b63323e597dcb4a --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_75_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:PPLCNet_x0_75 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..cfda57015d918b79bc3ab9103314b53803544c39 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x0_75 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_75.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_75.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x0_75.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_75_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_infer_python.txt index 14b5ec345c3d958efedcb1cd693f1a7c2d8254f4..3f16d3640e06a6f1fd87c4946e6d8d2efb4d9822 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x0_75_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PPLCNet/PPLCNet_x1_0_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/PPLCNet_x1_0_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..64369b742ff84ca123e6d562929a3221f828a97c --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x1_0_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:PPLCNet_x1_0 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..c335e54284af5eaeeb321960ba2b27b553a04ca4 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_infer_python.txt index 97bec1f6d88fe5568347e00297206bb557910e9f..f17667e285652efbeea24f974b55982541762bda 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x1_0_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PPLCNet/PPLCNet_x1_5_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/PPLCNet_x1_5_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..cbf5cde57cef28e620841ac25badb97fcc2198ab --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x1_5_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:PPLCNet_x1_5 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..6e170df68bd3d238e773744851795d62b09283f2 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_infer_python.txt index 6faaddb10fb6d7f4698d27b53eb70640cc46a82a..6a40f04f2d96697998ecc30e7a9b1fbffdfd1941 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x1_5_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PPLCNet/PPLCNet_x2_0_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/PPLCNet_x2_0_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..4382f79ed0bf3546b497aaa1da0dc19b7482d8fd --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x2_0_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:PPLCNet_x2_0 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b3a6f7b57ff8e40e68ee42a4ae6487a56096c0f9 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_infer_python.txt index 36145d52e4605dee21ba8af358f1bfa36c043b16..ff4939cec8b735ea7babaa6e23c5a4a25497158a 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x2_0_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PPLCNet/PPLCNet_x2_5_lite_arm_cpu_cpp.txt b/test_tipc/config/PPLCNet/PPLCNet_x2_5_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..7b5b0904b58f9361553233e82d8b3ac81e72faa3 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x2_5_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:PPLCNet_x2_5 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_amp_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..26e95cb7897d83a45355300bf1fdd4d1cf4aa3b5 --- /dev/null +++ b/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_amp_infer_python.txt @@ -0,0 +1,51 @@ +===========================train_params=========================== +model_name:PPLCNet_x2_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/PPLCNet/PPLCNet_x2_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null diff --git a/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_infer_python.txt b/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_infer_python.txt index f369e740559b99381dae0e13c69007aee4c576dc..723939b03fa0d6adb69534d24a69a1bc2c07bf36 100644 --- a/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_infer_python.txt +++ b/test_tipc/config/PPLCNet/PPLCNet_x2_5_train_infer_python.txt @@ -49,3 +49,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.save_log_path:null -o Global.benchmark:True null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PVTV2/PVT_V2_B0.txt b/test_tipc/config/PVTV2/PVT_V2_B0_train_infer_python.txt similarity index 94% rename from test_tipc/config/PVTV2/PVT_V2_B0.txt rename to test_tipc/config/PVTV2/PVT_V2_B0_train_infer_python.txt index 1e0a3800c06f9c26a2e1f04167e9fb9996ddee82..a4229d854a9706cfa9bd0baeb18c6c59803197ce 100644 --- a/test_tipc/config/PVTV2/PVT_V2_B0.txt +++ b/test_tipc/config/PVTV2/PVT_V2_B0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PVTV2/PVT_V2_B1.txt b/test_tipc/config/PVTV2/PVT_V2_B1_train_infer_python.txt similarity index 94% rename from test_tipc/config/PVTV2/PVT_V2_B1.txt rename to test_tipc/config/PVTV2/PVT_V2_B1_train_infer_python.txt index 489f8c80d363ba8ae6f98e141e3467e5b7eccb2c..5394790a03b68366a83b4ea2fba3e67c988d7f1e 100644 --- a/test_tipc/config/PVTV2/PVT_V2_B1.txt +++ b/test_tipc/config/PVTV2/PVT_V2_B1_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/PVTV2/PVT_V2_B2_Linear.txt b/test_tipc/config/PVTV2/PVT_V2_B2_Linear_train_infer_python.txt similarity index 78% rename from test_tipc/config/PVTV2/PVT_V2_B2_Linear.txt rename to test_tipc/config/PVTV2/PVT_V2_B2_Linear_train_infer_python.txt index 2fd530c97cea2233d73fb11f2dddebdc70e73980..f50107fea5309eee26b3e46f1d9f7388ff7f1b59 100644 --- a/test_tipc/config/PVTV2/PVT_V2_B2_Linear.txt +++ b/test_tipc/config/PVTV2/PVT_V2_B2_Linear_train_infer_python.txt @@ -13,7 +13,7 @@ train_infer_img_dir:./dataset/ILSVRC2012/val null:null ## trainer:norm_train -norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +norm_train:tools/train.py -c ppcls/configs/ImageNet/PVTV2/PVT_V2_B2_Linear.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o Global.print_batch_step=1 pact_train:null fpgm_train:null distill_train:null @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:128 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/PVTV2/PVT_V2_B2.txt b/test_tipc/config/PVTV2/PVT_V2_B2_train_infer_python.txt similarity index 93% rename from test_tipc/config/PVTV2/PVT_V2_B2.txt rename to test_tipc/config/PVTV2/PVT_V2_B2_train_infer_python.txt index 8d274cd20fcd9cbc2cfaa387f5acf1399330bc8c..d2de07223364b1eb78597c7de97e8a10725bbc18 100644 --- a/test_tipc/config/PVTV2/PVT_V2_B2.txt +++ b/test_tipc/config/PVTV2/PVT_V2_B2_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/PVTV2/PVT_V2_B3.txt b/test_tipc/config/PVTV2/PVT_V2_B3_train_infer_python.txt similarity index 93% rename from test_tipc/config/PVTV2/PVT_V2_B3.txt rename to test_tipc/config/PVTV2/PVT_V2_B3_train_infer_python.txt index f375fcf2189823bf91a89e87d7d67f7d7047cca9..8f71e148fdf61147d6b4c5acdd818e8051222409 100644 --- a/test_tipc/config/PVTV2/PVT_V2_B3.txt +++ b/test_tipc/config/PVTV2/PVT_V2_B3_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/PVTV2/PVT_V2_B4.txt b/test_tipc/config/PVTV2/PVT_V2_B4_train_infer_python.txt similarity index 93% rename from test_tipc/config/PVTV2/PVT_V2_B4.txt rename to test_tipc/config/PVTV2/PVT_V2_B4_train_infer_python.txt index f0c7084fb2f70fb030ea331e33731b9511fe005b..304deb6872641c7cb5ae51ccc1aa241fc621e660 100644 --- a/test_tipc/config/PVTV2/PVT_V2_B4.txt +++ b/test_tipc/config/PVTV2/PVT_V2_B4_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/PVTV2/PVT_V2_B5.txt b/test_tipc/config/PVTV2/PVT_V2_B5_train_infer_python.txt similarity index 93% rename from test_tipc/config/PVTV2/PVT_V2_B5.txt rename to test_tipc/config/PVTV2/PVT_V2_B5_train_infer_python.txt index 9383851c87fb4897cd5301ad279fdcdb678a02f1..9527cb2d385ed217e3212fd535629b2407ff8dee 100644 --- a/test_tipc/config/PVTV2/PVT_V2_B5.txt +++ b/test_tipc/config/PVTV2/PVT_V2_B5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/ReXNet/ReXNet_1_0_train_amp_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..9ae14e503dbd6a13a0eddf3a29bbbc13d3e73fd8 --- /dev/null +++ b/test_tipc/config/ReXNet/ReXNet_1_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ReXNet/ReXNet_1_0_train_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_1_0_train_infer_python.txt index 352ff5ec278d2db4a07bf1db4af8b8f040d200e6..4f505f6e283642df524cc1845e1e0ba553cb4b00 100644 --- a/test_tipc/config/ReXNet/ReXNet_1_0_train_infer_python.txt +++ b/test_tipc/config/ReXNet/ReXNet_1_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ReXNet/ReXNet_1_3_train_amp_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_1_3_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f5c7aed52a243a03ff63269c5312b021786a7d13 --- /dev/null +++ b/test_tipc/config/ReXNet/ReXNet_1_3_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_1_3 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_3.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_3.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_3.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_3_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ReXNet/ReXNet_1_3_train_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_1_3_train_infer_python.txt index 4c76badb85477c02f867b1d7d9b8808dffac57be..e21cfa98f7a8a3bdea58a5a7606d601f830d5a92 100644 --- a/test_tipc/config/ReXNet/ReXNet_1_3_train_infer_python.txt +++ b/test_tipc/config/ReXNet/ReXNet_1_3_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ReXNet/ReXNet_1_5_train_amp_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_1_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3399aad8da871d85b4b3c37acfcb657dcfa58c04 --- /dev/null +++ b/test_tipc/config/ReXNet/ReXNet_1_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ReXNet/ReXNet_1_5_train_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_1_5_train_infer_python.txt index b302911e07fa20580bb32e60b078978431646af8..4e750926ef4d5e933b4b05aba63f15dda9c9d7cb 100644 --- a/test_tipc/config/ReXNet/ReXNet_1_5_train_infer_python.txt +++ b/test_tipc/config/ReXNet/ReXNet_1_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ReXNet/ReXNet_2_0_train_amp_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_2_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f036e87b9c7671150f537ae3d458218e83748115 --- /dev/null +++ b/test_tipc/config/ReXNet/ReXNet_2_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ReXNet/ReXNet_2_0_train_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_2_0_train_infer_python.txt index c52d563ccc269db5ff6b150143558f3707c55d6e..b98182f27155e2a8d14a3d88781e86294f767aca 100644 --- a/test_tipc/config/ReXNet/ReXNet_2_0_train_infer_python.txt +++ b/test_tipc/config/ReXNet/ReXNet_2_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ReXNet/ReXNet_3_0_train_amp_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_3_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3263e6e16ad1a259e2804c40f6ac6a4319eca726 --- /dev/null +++ b/test_tipc/config/ReXNet/ReXNet_3_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ReXNet_3_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_3_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_3_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ReXNet/ReXNet_3_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ReXNet_3_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ReXNet/ReXNet_3_0_train_infer_python.txt b/test_tipc/config/ReXNet/ReXNet_3_0_train_infer_python.txt index a2047c4495bface98fc015ce4aa80f0ead8a12fe..d8b9ac8ed95322cfd6a3673815f63c31b73e0716 100644 --- a/test_tipc/config/ReXNet/ReXNet_3_0_train_infer_python.txt +++ b/test_tipc/config/ReXNet/ReXNet_3_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/RedNet/RedNet101_train_amp_infer_python.txt b/test_tipc/config/RedNet/RedNet101_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..8255130c48d2b5083c2e629d8a4381203b98806f --- /dev/null +++ b/test_tipc/config/RedNet/RedNet101_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet101 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet101.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet101.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet101.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet101_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/RedNet/RedNet101_train_infer_python.txt b/test_tipc/config/RedNet/RedNet101_train_infer_python.txt index 6d7d3b9d82d968ffbd2134fdbff2d8ac49221d4a..948f1fea96edb779f3cdd12867c5813e1a5e8b74 100644 --- a/test_tipc/config/RedNet/RedNet101_train_infer_python.txt +++ b/test_tipc/config/RedNet/RedNet101_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/RedNet/RedNet152_train_amp_infer_python.txt b/test_tipc/config/RedNet/RedNet152_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..5b208005b8357165d130d4e74e1cc943efe36b7c --- /dev/null +++ b/test_tipc/config/RedNet/RedNet152_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet152 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet152.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet152.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet152.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet152_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/RedNet/RedNet152_train_infer_python.txt b/test_tipc/config/RedNet/RedNet152_train_infer_python.txt index 82e8189d5b69858bc70f93cbfd9d22d2d9177b65..fca7c0e0e10d4d03bc85eb28f66a9b58dfd1a4e1 100644 --- a/test_tipc/config/RedNet/RedNet152_train_infer_python.txt +++ b/test_tipc/config/RedNet/RedNet152_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/RedNet/RedNet26_train_amp_infer_python.txt b/test_tipc/config/RedNet/RedNet26_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1aa6bbd4d0ca15b7b07852b73b8c4c58175542b4 --- /dev/null +++ b/test_tipc/config/RedNet/RedNet26_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet26 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet26.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet26.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet26.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet26_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/RedNet/RedNet26_train_infer_python.txt b/test_tipc/config/RedNet/RedNet26_train_infer_python.txt index 53e5ce00875a3c890de02c8125c621915e998f58..7b13978e71ad058dbcbceed695b1f8da0385abe8 100644 --- a/test_tipc/config/RedNet/RedNet26_train_infer_python.txt +++ b/test_tipc/config/RedNet/RedNet26_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/RedNet/RedNet38_train_amp_infer_python.txt b/test_tipc/config/RedNet/RedNet38_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..05715a860b2f21b30d5c869c1115d98dcf41cd84 --- /dev/null +++ b/test_tipc/config/RedNet/RedNet38_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet38 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet38.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet38.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet38.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet38_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/RedNet/RedNet38_train_infer_python.txt b/test_tipc/config/RedNet/RedNet38_train_infer_python.txt index 3ec247162616a1a4c28ce29f06a30b0b940a9806..9ba325fe0628d45d43759a1a3a01e7a9a8f10dc6 100644 --- a/test_tipc/config/RedNet/RedNet38_train_infer_python.txt +++ b/test_tipc/config/RedNet/RedNet38_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/RedNet/RedNet50_train_amp_infer_python.txt b/test_tipc/config/RedNet/RedNet50_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3ea2e0d316e239af4eab2f93d21b53e28d88fe12 --- /dev/null +++ b/test_tipc/config/RedNet/RedNet50_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:RedNet50 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/RedNet/RedNet50.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/RedNet/RedNet50.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/RedNet/RedNet50.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RedNet50_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/RedNet/RedNet50_train_infer_python.txt b/test_tipc/config/RedNet/RedNet50_train_infer_python.txt index bb36200ad1aab66f8bfe4d2efd2fb648972780cd..14265c9ff3e39b1792a9e035f6d74f3e92e5c8a7 100644 --- a/test_tipc/config/RedNet/RedNet50_train_infer_python.txt +++ b/test_tipc/config/RedNet/RedNet50_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_amp_infer_python.txt b/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..6a9f287429be801d96be984048bd0be5fedb4c68 --- /dev/null +++ b/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net101_vd_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net101_vd_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net101_vd_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net101_vd_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net101_vd_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_infer_python.txt b/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_infer_python.txt index f5542adc7da3732bb1a94f9c7193a153cc806260..c8bab72c10aedb0515ad5381cb4848a7785ece19 100644 --- a/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_infer_python.txt +++ b/test_tipc/config/Res2Net/Res2Net101_vd_26w_4s_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_amp_infer_python.txt b/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..c8157cae60a41cfa15458d4b697000cd2f9f20d0 --- /dev/null +++ b/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net200_vd_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net200_vd_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net200_vd_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net200_vd_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net200_vd_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_infer_python.txt b/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_infer_python.txt index 40d7f4103d626c5ec34b119f88b0c9b1dfd790f2..6fab68f036d76ef948621b253418cee0ca364bc6 100644 --- a/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_infer_python.txt +++ b/test_tipc/config/Res2Net/Res2Net200_vd_26w_4s_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_amp_infer_python.txt b/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..6bfee99062699c581a38bb2fe14ed6ad986750c2 --- /dev/null +++ b/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net50_14w_8s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_14w_8s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_14w_8s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_14w_8s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_14w_8s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_infer_python.txt b/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_infer_python.txt index c6300bc94f2c4798c90da8acd50634b3ebd857a9..f74ddf866772c2eb3c64a07136a4643f95ebdad0 100644 --- a/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_infer_python.txt +++ b/test_tipc/config/Res2Net/Res2Net50_14w_8s_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_amp_infer_python.txt b/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4a33c79c4c7f4b216e9974376fedc6024f269c8e --- /dev/null +++ b/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net50_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_infer_python.txt b/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_infer_python.txt index 7491525825dd71ed6a23a0314bb775b4fab8e09c..1c4c9242bb7affae0045936d65ded2b8228d2fb0 100644 --- a/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_infer_python.txt +++ b/test_tipc/config/Res2Net/Res2Net50_26w_4s_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_amp_infer_python.txt b/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..cd3409a333240e4026960a089c7720b97f772eba --- /dev/null +++ b/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Res2Net50_vd_26w_4s +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_vd_26w_4s.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_vd_26w_4s.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Res2Net/Res2Net50_vd_26w_4s.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Res2Net50_vd_26w_4s_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_infer_python.txt b/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_infer_python.txt index cbdc256373aba2d42b47b08c2f261b654d4fe71d..6e8b05f0fadbd9a3a6016631e2e57b75d539df70 100644 --- a/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_infer_python.txt +++ b/test_tipc/config/Res2Net/Res2Net50_vd_26w_4s_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_amp_infer_python.txt b/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0a8e0d0a97e8fbb24579e0d87f720169f1d5e722 --- /dev/null +++ b/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeSt50_fast_1s1x64d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50_fast_1s1x64d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50_fast_1s1x64d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50_fast_1s1x64d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_fast_1s1x64d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_infer_python.txt b/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_infer_python.txt index 66ce66a559d120f54af2df637b7077929c841b41..00166a60ca3a70b16d38a6333f49261cd3a58347 100644 --- a/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_infer_python.txt +++ b/test_tipc/config/ResNeSt/ResNeSt50_fast_1s1x64d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeSt/ResNeSt50_train_amp_infer_python.txt b/test_tipc/config/ResNeSt/ResNeSt50_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1d0e59216c17dacaf255f3ce7a4d3020d4d48c48 --- /dev/null +++ b/test_tipc/config/ResNeSt/ResNeSt50_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeSt50 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeSt/ResNeSt50.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeSt50_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeSt/ResNeSt50_train_infer_python.txt b/test_tipc/config/ResNeSt/ResNeSt50_train_infer_python.txt index 2d1bfb2a24a3fcdca2e6d6596956990e2c566d36..70681273249675b16b04b370333b396d7f5598aa 100644 --- a/test_tipc/config/ResNeSt/ResNeSt50_train_infer_python.txt +++ b/test_tipc/config/ResNeSt/ResNeSt50_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1ca28d104c9b4f9d1e43869c4e3fb4b1cf6c0019 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_infer_python.txt index eb21af69f0b98ec97ee7b34a1e74980c60d23a89..4f9df21ec66d409b84a6ac3a7a6952afec168c00 100644 --- a/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt101_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..69db3f17bf2b6e142cbeaecf4f5e131571966676 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_infer_python.txt index 8e4ac811456d6738a6d5988349a5a74bffcfa7f3..e15b752694e7b438cd8964a647e9771d73c9b61d 100644 --- a/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt101_64x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..22edd3b99ed6355199d1cfcf13f481be3d76cc04 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_infer_python.txt index cf9211f609f51f67da39d04272dfbe97679d0346..87d8396c7d993e353f64fa1a56742de308a906a9 100644 --- a/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt101_vd_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..73f786f113a566bde9a1fb75e49b223c6701756d --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_vd_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_vd_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_vd_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_infer_python.txt index eae4d0e4a7d5f1782ca73244f5043a1ac9216496..4ec192cc76656532a481436aa48e18b49d3be296 100644 --- a/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt101_vd_64x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1ca28d104c9b4f9d1e43869c4e3fb4b1cf6c0019 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt101_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt101_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt101_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_infer_python.txt index eb21af69f0b98ec97ee7b34a1e74980c60d23a89..4f9df21ec66d409b84a6ac3a7a6952afec168c00 100644 --- a/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt152_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3841f2b7e72e3d6c82a879e833f65758a2f4a934 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt152_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_infer_python.txt index c8a85f6fc85e1815ad30195626b373dd8119c8f7..0d1bdf8f1a497adafd12ab1176eb1e32b823309f 100644 --- a/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt152_64x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b0baf65150176639f60200ebd73b3b708aff41ea --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt152_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_infer_python.txt index 4abaa874c6abd87025a60aac8378151ac9935fb6..2cb58880278f9edd85255d1dcfd4260d475fdfd7 100644 --- a/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt152_vd_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..573b579ff161aca2741867bece064ff10fd5bc22 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt152_vd_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt152_vd_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt152_vd_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_infer_python.txt index 8d8a4ea190dac6fe38e698e91ae2919cb7972331..20765c423704335a60ea56aab615979349e1a036 100644 --- a/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt152_vd_64x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b28bd26fc7f2af8996722593f41129d3898b3e93 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_infer_python.txt index 4cb942f13b53b72c4b90fab2da584a84f3e9dab1..d1fcf270c9da80660800eec0cec9d24599fac7f4 100644 --- a/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt50_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..9d04710ab9b93546f4ff235d48c95eaf0cf25e40 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_infer_python.txt index 0cad348ce7c2a76211bb36bf433e8e89fc5a9146..81eab9aa344493d20e19a79e3458b2988475717c 100644 --- a/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt50_64x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b502731f745810f6108c697a8b082e4805a55a89 --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_infer_python.txt index fc7456019c2fe7bdc000a5a125276fc83deb9c32..1d18804f41c09d6f09959d5efb979cee9b81242e 100644 --- a/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt50_vd_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_amp_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b0b0b3fe272ca5254fe59faa504cb3905b8c22fc --- /dev/null +++ b/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNeXt50_vd_64x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_64x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_64x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNeXt/ResNeXt50_vd_64x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNeXt50_vd_64x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_infer_python.txt b/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_infer_python.txt index b2817dbad3d2c5481f9331eb57216748aba37950..b63ed3afd399ed233a470436a8849205fdb982e3 100644 --- a/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_infer_python.txt +++ b/test_tipc/config/ResNeXt/ResNeXt50_vd_64x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet101_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet101_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..9172ef8cf54d76ffe17079a17b6bf796df2a3351 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet101_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet101 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet101.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet101.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet101.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet101_train_infer_python.txt b/test_tipc/config/ResNet/ResNet101_train_infer_python.txt index f10760d7438afe83b0bf313572a53d9f18c32e6d..aee834cff7f5a2e86c798f1974dd85f32745143f 100644 --- a/test_tipc/config/ResNet/ResNet101_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet101_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet101_vd_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet101_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..eb8238584b3828a12c7c8d75118dcc78da8a0562 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet101_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet101_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet101_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet101_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet101_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet101_vd_train_infer_python.txt b/test_tipc/config/ResNet/ResNet101_vd_train_infer_python.txt index 7027d30c557c91757225baebe38a9a03dd9221fb..3c9745b35f06a8f3485c61706742cab79d53bb2a 100644 --- a/test_tipc/config/ResNet/ResNet101_vd_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet101_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet152_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet152_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..8549cc07cde50b4f51da729aacebed08470c2187 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet152_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet152 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet152.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +to_static_train:-o Global.to_static=True +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet152.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet152.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet152_train_infer_python.txt b/test_tipc/config/ResNet/ResNet152_train_infer_python.txt index 1d4cc80d92c0ea1aabf2a6096e0c657e28eaeb75..e45266a9a83dfd93561c91e3543f668d802cc440 100644 --- a/test_tipc/config/ResNet/ResNet152_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet152_train_infer_python.txt @@ -17,7 +17,7 @@ norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet152.yaml -o Glo pact_train:null fpgm_train:null distill_train:null -null:null +to_static_train:-o Global.to_static=True null:null ## ===========================eval_params=========================== @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:32 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet152_vd_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet152_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a8e8c630516cd8ce1c5918a60e0feffdfd746270 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet152_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet152_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet152_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet152_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet152_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet152_vd_train_infer_python.txt b/test_tipc/config/ResNet/ResNet152_vd_train_infer_python.txt index 409132861cabeaa3dd18d5be918747f4ab5f441b..0ca10de1d62986dc8c2f1ebe1cb84cc59620fd87 100644 --- a/test_tipc/config/ResNet/ResNet152_vd_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet152_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet18_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet18_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4f2badaf2ce2f61ffe7676825411fdb0f955cad8 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet18_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet18 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet18.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet18.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet18.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet18_train_infer_python.txt b/test_tipc/config/ResNet/ResNet18_train_infer_python.txt index 06d550499075e9660ed5de2acafadb65c190f5bb..c71707d31a9ea0c440c0c05536d907741fc77b5f 100644 --- a/test_tipc/config/ResNet/ResNet18_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet18_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet18_vd_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet18_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4dc479177c33e078023dcd352cba51b74b5417b2 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet18_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet18_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet18_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet18_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet18_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet18_vd_train_infer_python.txt b/test_tipc/config/ResNet/ResNet18_vd_train_infer_python.txt index d3d2bcf1dcb2ee76dfee377867566fde9381c332..c9ba57d2532a55de69f8b507985af7f03671072a 100644 --- a/test_tipc/config/ResNet/ResNet18_vd_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet18_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet200_vd_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet200_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a829b20fa8c5588eaea68cf705bf06fd6fd5fef7 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet200_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet200_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet200_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet200_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet200_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet200_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet200_vd_train_infer_python.txt b/test_tipc/config/ResNet/ResNet200_vd_train_infer_python.txt index e742e87ad4e1a57518a66d1e8e00fb11591cb2af..810aef1f402ac9caab76dd931db7ee39de57e885 100644 --- a/test_tipc/config/ResNet/ResNet200_vd_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet200_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet34_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet34_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..301ca3d91f6130a49b66ba551e9cabe8a426d0b1 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet34_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet34 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet34.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet34.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet34.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet34_train_infer_python.txt b/test_tipc/config/ResNet/ResNet34_train_infer_python.txt index 4a520b55d9306f2f3cdd008a7cbb44ec5ec7363f..7f1fef2962be1f33b7638793aec9197ae8817c2e 100644 --- a/test_tipc/config/ResNet/ResNet34_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet34_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet34_vd_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet34_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..62473756e7f8b5195a5782e639254643a44444ce --- /dev/null +++ b/test_tipc/config/ResNet/ResNet34_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet34_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet34_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet34_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet34_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet34_vd_train_infer_python.txt b/test_tipc/config/ResNet/ResNet34_vd_train_infer_python.txt index b956c83dc3a3aa89970a73b42c83a500ca7e1ae8..4771fa0bb1baf06fad9db5aa2d663c4d503f4f3f 100644 --- a/test_tipc/config/ResNet/ResNet34_vd_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet34_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet50_lite_arm_cpu_cpp.txt b/test_tipc/config/ResNet/ResNet50_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..a896023568bfd3ace5026749c170b640662c62b1 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet50_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:ResNet50 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/ResNet/ResNet50_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet50_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a398086aaf466d91e330d0e794943324e9913870 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet50_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +to_static_train:-o Global.to_static=True +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet50_train_infer_python.txt b/test_tipc/config/ResNet/ResNet50_train_infer_python.txt index 6775f9528f38078cabc33c10c6d4d48168f0061d..6a668ccd311c82f33f2843ac1ee086027635dd0d 100644 --- a/test_tipc/config/ResNet/ResNet50_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet50_train_infer_python.txt @@ -17,7 +17,7 @@ norm_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50.yaml -o Glob pact_train:null fpgm_train:null distill_train:null -null:null +to_static_train:-o Global.to_static=True null:null ## ===========================eval_params=========================== @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:128 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ResNet/ResNet50_vd_FPGM_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet50_vd_FPGM_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0b00effeb535ecefd85507f869815dc5cde1253a --- /dev/null +++ b/test_tipc/config/ResNet/ResNet50_vd_FPGM_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50_vd_FPGM +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet50_vd_PACT_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet50_vd_PACT_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..e4c65b010a1dd36f4a5b34cf1d03980913c11c18 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet50_vd_PACT_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50_vd_PACT +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet50_vd_lite_arm_cpu_cpp.txt b/test_tipc/config/ResNet/ResNet50_vd_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..3568c0ae15d60c46896665bfad6e1019ec55c077 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet50_vd_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:ResNet50_vd +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/ResNet/ResNet50_vd_train_amp_infer_python.txt b/test_tipc/config/ResNet/ResNet50_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a9cc2d027752db2880ff1655b69c1adc7d82bb19 --- /dev/null +++ b/test_tipc/config/ResNet/ResNet50_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ResNet50_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml +quant_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_quantization.yaml +fpgm_export:tools/export_model.py -c ppcls/configs/slim/ResNet50_vd_prune.yaml +distill_export:null +kl_quant:deploy/slim/quant_post_static.py -c ppcls/configs/ImageNet/ResNet/ResNet50_vd.yaml -o Global.save_inference_dir=./inference +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt b/test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt index d16ecd3fb7e7cdc950ba22691d3406f5d1311d8c..bf4de414faa0dce0bdd3894b12964b0d0b762f35 100644 --- a/test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt +++ b/test_tipc/config/ResNet/ResNet50_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SENet/SENet154_vd_train_amp_infer_python.txt b/test_tipc/config/SENet/SENet154_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..2bd7b3c5bf2c8ec7784268ece10e3f35c7c199ee --- /dev/null +++ b/test_tipc/config/SENet/SENet154_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SENet154_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SENet154_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SENet154_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SENet154_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SENet154_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SENet/SENet154_vd_train_infer_python.txt b/test_tipc/config/SENet/SENet154_vd_train_infer_python.txt index 29d29547805852c43f1416588c4137936730d278..5da5e4bca6b4b8decdef7fb9b4b3c16eefcd0a50 100644 --- a/test_tipc/config/SENet/SENet154_vd_train_infer_python.txt +++ b/test_tipc/config/SENet/SENet154_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_amp_infer_python.txt b/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..57afa0d0a207fee2a4d5029205eb13d4154a86ae --- /dev/null +++ b/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNeXt101_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt101_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt101_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_infer_python.txt b/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_infer_python.txt index 023dceba669493ebe891cdb5e3f7a13c002dacb8..1a61d0c229da34b3ca605dc4019908533c82fc9b 100644 --- a/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_infer_python.txt +++ b/test_tipc/config/SENet/SE_ResNeXt101_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_amp_infer_python.txt b/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ae0a334de05c91a628099674da99d9bf0d73fc29 --- /dev/null +++ b/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNeXt50_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_infer_python.txt b/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_infer_python.txt index e6c6db7983a127b1f66d76c58456e322c0fe7a31..a8bd8b553a4905ebc8e7e7ca9ae22aaaeb52512a 100644 --- a/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_infer_python.txt +++ b/test_tipc/config/SENet/SE_ResNeXt50_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_amp_infer_python.txt b/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..3cea2c6afdf16bd8e05d5a8e3a7b3d464149402d --- /dev/null +++ b/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNeXt50_vd_32x4d +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_vd_32x4d.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_vd_32x4d.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNeXt50_vd_32x4d.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNeXt50_vd_32x4d_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_infer_python.txt b/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_infer_python.txt index c2704f9979b675eebeb3080f507ba20e95953c98..1ea4449d5cd250534fad7492ff73daddd7dd3b74 100644 --- a/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_infer_python.txt +++ b/test_tipc/config/SENet/SE_ResNeXt50_vd_32x4d_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SENet/SE_ResNet18_vd_train_amp_infer_python.txt b/test_tipc/config/SENet/SE_ResNet18_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..872e1ce72d015723c570a19e7ca849631e050e36 --- /dev/null +++ b/test_tipc/config/SENet/SE_ResNet18_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNet18_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNet18_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNet18_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNet18_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet18_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SENet/SE_ResNet18_vd_train_infer_python.txt b/test_tipc/config/SENet/SE_ResNet18_vd_train_infer_python.txt index 095b1b98322b34fbed1117defb47f6edbf0a9179..1a0f673efb9c695ef507532fde00f4a344ac3c87 100644 --- a/test_tipc/config/SENet/SE_ResNet18_vd_train_infer_python.txt +++ b/test_tipc/config/SENet/SE_ResNet18_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SENet/SE_ResNet34_vd_train_amp_infer_python.txt b/test_tipc/config/SENet/SE_ResNet34_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4ac5c06b4334a727bd8981cb6fd54e0af4d5b413 --- /dev/null +++ b/test_tipc/config/SENet/SE_ResNet34_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNet34_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNet34_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNet34_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNet34_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet34_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SENet/SE_ResNet34_vd_train_infer_python.txt b/test_tipc/config/SENet/SE_ResNet34_vd_train_infer_python.txt index 260bc637fcaa21b134b64754a80295360b5485c8..742dd4ba96133338bfce9a56433109eec2e38808 100644 --- a/test_tipc/config/SENet/SE_ResNet34_vd_train_infer_python.txt +++ b/test_tipc/config/SENet/SE_ResNet34_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SENet/SE_ResNet50_vd_train_amp_infer_python.txt b/test_tipc/config/SENet/SE_ResNet50_vd_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..d9bc19b579704acce852c651d60d6410596ac497 --- /dev/null +++ b/test_tipc/config/SENet/SE_ResNet50_vd_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SE_ResNet50_vd +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SENet/SE_ResNet50_vd.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SENet/SE_ResNet50_vd.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SENet/SE_ResNet50_vd.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SE_ResNet50_vd_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SENet/SE_ResNet50_vd_train_infer_python.txt b/test_tipc/config/SENet/SE_ResNet50_vd_train_infer_python.txt index b486f3c6c860b12ab888b031e38830903beda188..9edf4dd24fbd819cd40ef2cebd302f5f9114ae5e 100644 --- a/test_tipc/config/SENet/SE_ResNet50_vd_train_infer_python.txt +++ b/test_tipc/config/SENet/SE_ResNet50_vd_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_amp_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..9822328e323c36d56a7fec860f6a55474c446a81 --- /dev/null +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_swish +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_swish.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_swish.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_swish.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_infer_python.txt index 616e269ccda2b9edc48edd1c8d73d108a5d0a761..da05d8a3af323b1799128ca70124a9b872c180b6 100644 --- a/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_infer_python.txt +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_swish_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_amp_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..e2e6a4c995ebc9739328a8dc3b83f17a4545633b --- /dev/null +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_25 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_25.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_infer_python.txt index df208125c6fbbf088da9dec15239bd0a7a66da18..1dbc646c83e6794bb117dff17e23df4a94577a94 100644 --- a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_infer_python.txt +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_25_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_amp_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..97b59cc65db4a81779a9429961858ecc41790cc3 --- /dev/null +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_33 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_33.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_infer_python.txt index 91df1ac50c0d1f03f36e2838eda2757705aaace6..3c429924af3d589088ab7ee57a6052d6b241d6c9 100644 --- a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_infer_python.txt +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_33_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_amp_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..4dc126ab690428a4ba9af8cda59ffaf445dc0021 --- /dev/null +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x0_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x0_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_infer_python.txt index e0ee753217894dc99961a433cc495bc6d387e19d..ac890e7ab2ec563ecc53155a23816f8dc7eae373 100644 --- a/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_infer_python.txt +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x0_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_amp_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..e738f37653716ca6a06764c80645ee0e200a8ce7 --- /dev/null +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt index a0f354c01c0df8be54ca1495f75cb5f7aacf6078..9b3cc1d5af028a627f99fccdb7a4a8a3dc11e2cc 100644 --- a/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt @@ -17,7 +17,7 @@ norm_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_0 pact_train:null fpgm_train:null distill_train:null -null:null +to_static_train:-o Global.to_static=True null:null ## ===========================eval_params=========================== @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:256|1536 +fp_items:fp32 +epoch:2 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_amp_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..7317c4780f60296247cf96d2753e7e761328fd39 --- /dev/null +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x1_5 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_5.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_5.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x1_5.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_infer_python.txt index f9ab9c15d5f59479509bd13adb898b9dd9cbed88..3564be4dcd9e3ad9cd445e89373933a58ccff936 100644 --- a/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_infer_python.txt +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x1_5_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_amp_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..020b44b68fd5e6ae25270097a6ce5f3126ae5dd5 --- /dev/null +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ShuffleNetV2_x2_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x2_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x2_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/ShuffleNet/ShuffleNetV2_x2_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_infer_python.txt b/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_infer_python.txt index eea768a3af15cef6cb712eaec0f702834773d105..87b2b340c8044e95c05cbcb485860338811050c9 100644 --- a/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_infer_python.txt +++ b/test_tipc/config/ShuffleNet/ShuffleNetV2_x2_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_amp_infer_python.txt b/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..33746bcd1b837a67dcc69364a93d7353fcf0cd95 --- /dev/null +++ b/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SqueezeNet1_0 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_0.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_0.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_0.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_0_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_infer_python.txt b/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_infer_python.txt index 49c59a9e5df7503cbaccd0d68e2b52ded8252506..274d759db7cfbed50c584a367160280a38820940 100644 --- a/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_infer_python.txt +++ b/test_tipc/config/SqueezeNet/SqueezeNet1_0_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_amp_infer_python.txt b/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..943a05e46808bbc60acb956899e710c2fba9710c --- /dev/null +++ b/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SqueezeNet1_1 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_1.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_1.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SqueezeNet/SqueezeNet1_1.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SqueezeNet1_1_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_infer_python.txt b/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_infer_python.txt index 1b957af79626f5497861dabf5ed613943b2cf753..06ad8c4cce46d41ac9577d87cb763b175f303bf5 100644 --- a/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_infer_python.txt +++ b/test_tipc/config/SqueezeNet/SqueezeNet1_1_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_amp_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..d4e0a4122e027329a88c3890e5adea5f684f8123 --- /dev/null +++ b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_base_patch4_window12_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window12_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window12_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_infer_python.txt index e7161c8d0fcac3ad2b7363b22c6ec15c2f0d57da..45e4d8534f3e172899d619bcfcb19cdf1916dc65 100644 --- a/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_infer_python.txt +++ b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window12_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] \ No newline at end of file diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_amp_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..8e686d07753b1ed2a9f441f7417908196518e24a --- /dev/null +++ b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_base_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_base_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_base_patch4_window7_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_infer_python.txt index 2307086a746cf43daaa0f2862b623b3213fa8de3..4270e32ec27ca1cd233da7e20664b1d46575f3cf 100644 --- a/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_infer_python.txt +++ b/test_tipc/config/SwinTransformer/SwinTransformer_base_patch4_window7_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_amp_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..6ed1e29b0fb27171f4745c223e471e0b9d9a5153 --- /dev/null +++ b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_large_patch4_window12_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window12_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window12_384_22kto1k_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_infer_python.txt index f1a1873d99e5f85ae7264f36f8068f78d3d5dfbd..737c4d3bd7bfe3537170226361ce775a73e4f88d 100644 --- a/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_infer_python.txt +++ b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window12_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] \ No newline at end of file diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_amp_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..46434bdb9c7b9023eec8a8a426ea1ec9e9a21cf6 --- /dev/null +++ b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_large_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_large_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_large_patch4_window7_224_22kto1k_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_infer_python.txt index 359899e1a1094dc9c58cb3b5ca47fd66a7989765..7c45933a977d42434982d12d261bd82502d493b5 100644 --- a/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_infer_python.txt +++ b/test_tipc/config/SwinTransformer/SwinTransformer_large_patch4_window7_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_amp_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..0475fa099927aaccdfd3c039d06606e173f445a5 --- /dev/null +++ b/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_small_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_small_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_small_patch4_window7_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_infer_python.txt index 7f9edb81831ea1d7d798838382d692b1783a610d..1c862375db536c57eb5d648277b313f369f8a301 100644 --- a/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_infer_python.txt +++ b/test_tipc/config/SwinTransformer/SwinTransformer_small_patch4_window7_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_lite_arm_cpu_cpp.txt b/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_lite_arm_cpu_cpp.txt new file mode 100644 index 0000000000000000000000000000000000000000..ae84567f047d5c64f8e628af9509c6afca840aa7 --- /dev/null +++ b/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_lite_arm_cpu_cpp.txt @@ -0,0 +1,8 @@ +runtime_device:arm_cpu +lite_arm_work_path:/data/local/tmp/arm_cpu/ +lite_arm_so_path:inference_lite_lib.android.armv8/cxx/lib/libpaddle_light_api_shared.so +clas_model_file:SwinTransformer_tiny_patch4_window7_224 +inference_cmd:clas_system config.txt tabby_cat.jpg +--num_threads_list:1 +--batch_size_list:1 +--precision_list:FP32 diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_amp_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..aabba1e440395287fd4d1711a8020a72ac4f4ece --- /dev/null +++ b/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:SwinTransformer_tiny_patch4_window7_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/SwinTransformer_tiny_patch4_window7_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt b/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt index 80b76ed6536dd33711d81cfdd688e793ffa55db3..8d8d4b314d03dc98d095f137626592a33450bab2 100644 --- a/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt +++ b/test_tipc/config/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:64|104 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/TNT/TNT_small_train_amp_infer_python.txt b/test_tipc/config/TNT/TNT_small_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..1966ac184caecd85cb1475bfddad368fbb1ac085 --- /dev/null +++ b/test_tipc/config/TNT/TNT_small_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:TNT_small +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/TNT/TNT_small.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/TNT/TNT_small.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/TNT/TNT_small.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/TNT_small_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/TNT/TNT_small_train_infer_python.txt b/test_tipc/config/TNT/TNT_small_train_infer_python.txt index 20d4264836ada0efefd420c175c44c274f724ad5..1e99ce2d6d5f9da79c6c57e75e88f8b1d518101d 100644 --- a/test_tipc/config/TNT/TNT_small_train_infer_python.txt +++ b/test_tipc/config/TNT/TNT_small_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Twins/alt_gvt_base_train_amp_infer_python.txt b/test_tipc/config/Twins/alt_gvt_base_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..e14b8ddc9ab1fc9a1605057afa8c3905d47814f9 --- /dev/null +++ b/test_tipc/config/Twins/alt_gvt_base_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:alt_gvt_base +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/alt_gvt_base.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_base_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Twins/alt_gvt_base_train_infer_python.txt b/test_tipc/config/Twins/alt_gvt_base_train_infer_python.txt index a08a7326d8f302fdd030ec1e262c21ea7a6570f3..ef4b566c4627de9f9aee6fe7412ca6646e77ef7a 100644 --- a/test_tipc/config/Twins/alt_gvt_base_train_infer_python.txt +++ b/test_tipc/config/Twins/alt_gvt_base_train_infer_python.txt @@ -50,3 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================train_benchmark_params========================== +batch_size:64|144 +fp_items:fp32 +epoch:1 +--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile +flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096 +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/Twins/alt_gvt_large_train_amp_infer_python.txt b/test_tipc/config/Twins/alt_gvt_large_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..436bb75dabd38c2d66ff93cac9244467bd2e7525 --- /dev/null +++ b/test_tipc/config/Twins/alt_gvt_large_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:alt_gvt_large +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/alt_gvt_large.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_large_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Twins/alt_gvt_large_train_infer_python.txt b/test_tipc/config/Twins/alt_gvt_large_train_infer_python.txt index d75f8a14cba0cf347f6219cf288a8bbc903a8a32..271e0c2c079811c19df88f42e00454d03192772b 100644 --- a/test_tipc/config/Twins/alt_gvt_large_train_infer_python.txt +++ b/test_tipc/config/Twins/alt_gvt_large_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Twins/alt_gvt_small_train_amp_infer_python.txt b/test_tipc/config/Twins/alt_gvt_small_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f82314ba6e1d7283b0e5f49da739e332768d8a74 --- /dev/null +++ b/test_tipc/config/Twins/alt_gvt_small_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:alt_gvt_small +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/alt_gvt_small.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/alt_gvt_small_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Twins/alt_gvt_small_train_infer_python.txt b/test_tipc/config/Twins/alt_gvt_small_train_infer_python.txt index d0686e5d19efd2e8d921f9540cbb1f285861af88..26307ad1279df64e56a7647ecdafd7041708b3ac 100644 --- a/test_tipc/config/Twins/alt_gvt_small_train_infer_python.txt +++ b/test_tipc/config/Twins/alt_gvt_small_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Twins/pcpvt_base_train_amp_infer_python.txt b/test_tipc/config/Twins/pcpvt_base_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..cf85959c01a2c50f5d2636585d388d3a3fa966a2 --- /dev/null +++ b/test_tipc/config/Twins/pcpvt_base_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:pcpvt_base +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Twins/pcpvt_base.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/pcpvt_base.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/pcpvt_base.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_base_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Twins/pcpvt_base_train_infer_python.txt b/test_tipc/config/Twins/pcpvt_base_train_infer_python.txt index 76486b819d9f8458c346ffbef937cc285dc37268..a1b3c0f65f487b7b96865df412de58b13c689fcb 100644 --- a/test_tipc/config/Twins/pcpvt_base_train_infer_python.txt +++ b/test_tipc/config/Twins/pcpvt_base_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Twins/pcpvt_large_train_amp_infer_python.txt b/test_tipc/config/Twins/pcpvt_large_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..622a62991c3c3332b18ac14238c3320b1ee7441d --- /dev/null +++ b/test_tipc/config/Twins/pcpvt_large_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:pcpvt_large +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Twins/pcpvt_large.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/pcpvt_large.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/pcpvt_large.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_large_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Twins/pcpvt_large_train_infer_python.txt b/test_tipc/config/Twins/pcpvt_large_train_infer_python.txt index 928c0b30db0dbc0e33108b7ee50cf91ddd84f821..275801c5edd8814e3ec54d35e82389de2c7afcd5 100644 --- a/test_tipc/config/Twins/pcpvt_large_train_infer_python.txt +++ b/test_tipc/config/Twins/pcpvt_large_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Twins/pcpvt_small_train_amp_infer_python.txt b/test_tipc/config/Twins/pcpvt_small_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..6d84c8f5eb10a53ea052fa7af303703b59235750 --- /dev/null +++ b/test_tipc/config/Twins/pcpvt_small_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:pcpvt_small +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Twins/pcpvt_small.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Twins/pcpvt_small.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Twins/pcpvt_small.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/pcpvt_small_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Twins/pcpvt_small_train_infer_python.txt b/test_tipc/config/Twins/pcpvt_small_train_infer_python.txt index 9ce13dfbbe127cf8b70fbba15918767f849cc607..95c4d54c3dcad9a98e446088f68f4a788474aca1 100644 --- a/test_tipc/config/Twins/pcpvt_small_train_infer_python.txt +++ b/test_tipc/config/Twins/pcpvt_small_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/VAN/VAN_tiny_train_infer_python.txt b/test_tipc/config/VAN/VAN_tiny_train_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..7d884d27ea666fa89fddd9487909d1371436e760 --- /dev/null +++ b/test_tipc/config/VAN/VAN_tiny_train_infer_python.txt @@ -0,0 +1,54 @@ +===========================train_params=========================== +model_name:VAN_tiny +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:norm_train +norm_train:tools/train.py -c ppcls/configs/ImageNet/VAN/VAN_tiny.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VAN/VAN_tiny.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VAN/VAN_tiny.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +inference_dir:null +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=248 -o PreProcess.transform_ops.2.NormalizeImage.mean=[0.5,0.5,0.5] -o PreProcess.transform_ops.2.NormalizeImage.std=[0.5,0.5,0.5] +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] diff --git a/test_tipc/config/VGG/VGG11_train_amp_infer_python.txt b/test_tipc/config/VGG/VGG11_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..cf0fe60c48e929ffbe4e188c54c557180dabf649 --- /dev/null +++ b/test_tipc/config/VGG/VGG11_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG11 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG11.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG11.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG11.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG11_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VGG/VGG11_train_infer_python.txt b/test_tipc/config/VGG/VGG11_train_infer_python.txt index 4d5f7b1096a210cbdec7bca268d40f47b09551dc..ed88fa4626640895b7ace894986513290f4cf344 100644 --- a/test_tipc/config/VGG/VGG11_train_infer_python.txt +++ b/test_tipc/config/VGG/VGG11_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/VGG/VGG13_train_amp_infer_python.txt b/test_tipc/config/VGG/VGG13_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..083a690c264a3441bafce83e4f9e9db398f17c29 --- /dev/null +++ b/test_tipc/config/VGG/VGG13_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG13 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG13.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG13.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG13.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG13_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VGG/VGG13_train_infer_python.txt b/test_tipc/config/VGG/VGG13_train_infer_python.txt index 737a56bb470a1d8291867c1a66c52c2b477e8645..bc04ce858352e82de2e1c7ca45e2121c666002fa 100644 --- a/test_tipc/config/VGG/VGG13_train_infer_python.txt +++ b/test_tipc/config/VGG/VGG13_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/VGG/VGG16_train_amp_infer_python.txt b/test_tipc/config/VGG/VGG16_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..30b422af895cbec8cbc89f05090e058609e04342 --- /dev/null +++ b/test_tipc/config/VGG/VGG16_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG16 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG16.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG16.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG16.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG16_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VGG/VGG16_train_infer_python.txt b/test_tipc/config/VGG/VGG16_train_infer_python.txt index b8bf46f03352fafad8b9e1025410c5092eda9a16..ff913c5352087f84289d0275c80763e30760cb00 100644 --- a/test_tipc/config/VGG/VGG16_train_infer_python.txt +++ b/test_tipc/config/VGG/VGG16_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/VGG/VGG19_train_amp_infer_python.txt b/test_tipc/config/VGG/VGG19_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ccbce1489852ce36701c7690abeaffe3e17221c5 --- /dev/null +++ b/test_tipc/config/VGG/VGG19_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:VGG19 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VGG/VGG19.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VGG/VGG19.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VGG/VGG19.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/VGG19_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VGG/VGG19_train_infer_python.txt b/test_tipc/config/VGG/VGG19_train_infer_python.txt index 3f5a8f3efa9a03fa88844c5a9ae4332648a6539f..9cd5ae5c691ab46690de7e0740c192bd602d4f84 100644 --- a/test_tipc/config/VGG/VGG19_train_infer_python.txt +++ b/test_tipc/config/VGG/VGG19_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_amp_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..74155b1f9215aebdb98ceb2f25dd430b462cfa82 --- /dev/null +++ b/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_base_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_infer_python.txt index d9f6e2f9a1d620cded9ccdbf76d3dcd9a3696926..bb73849e61cd2a8285d91f3db71c0796a510e3e7 100644 --- a/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_infer_python.txt +++ b/test_tipc/config/VisionTransformer/ViT_base_patch16_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_amp_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..aa23031bf79f193c060451e7615c670f31f834b0 --- /dev/null +++ b/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_base_patch16_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch16_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch16_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_infer_python.txt index ed88b5154410109ca114718542e5317627848bc5..20d3b73b94b59ad86878ad26ad1b9e10e401931c 100644 --- a/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_infer_python.txt +++ b/test_tipc/config/VisionTransformer/ViT_base_patch16_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] \ No newline at end of file diff --git a/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_amp_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..a5370b670ff3bcc0b2020f52f1e0ad8bbfd10850 --- /dev/null +++ b/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_base_patch32_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch32_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch32_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_base_patch32_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_base_patch32_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_infer_python.txt index 9c3abcdc0bef583284c8983cdc4e5a90b16cc1d9..f11d689651ba7f043c8c77eb9563bdc6d5b62320 100644 --- a/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_infer_python.txt +++ b/test_tipc/config/VisionTransformer/ViT_base_patch32_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] \ No newline at end of file diff --git a/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_amp_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..b86428b9fe077bcaeeff2ff6aa35161c977d141a --- /dev/null +++ b/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_large_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_infer_python.txt index ec77af6defbc0c062bac5517227890ea9f168909..0e07b743ec579e1a8a42c47eda89cc305aff67f4 100644 --- a/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_infer_python.txt +++ b/test_tipc/config/VisionTransformer/ViT_large_patch16_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_amp_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..c733d7e8274fc08dd782351397c65884a63d8763 --- /dev/null +++ b/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_large_patch16_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch16_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch16_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_infer_python.txt index e05c6df96bd61f67a4bf5e08e76f8cd5d334ddd5..06555842e7e2c37e70749836d9232348ecb1c18a 100644 --- a/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_infer_python.txt +++ b/test_tipc/config/VisionTransformer/ViT_large_patch16_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] \ No newline at end of file diff --git a/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_amp_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f6ead75b81e298c9684d2b1847356203afce3f26 --- /dev/null +++ b/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_large_patch32_384 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:2 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch32_384.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch32_384.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_large_patch32_384.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_large_patch32_384_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=384 -o PreProcess.transform_ops.1.CropImage.size=384 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_infer_python.txt index 43b313f58c047b94af2a64d7460910530bb488cf..e513814b7a995ce13805283841769005ed7376c6 100644 --- a/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_infer_python.txt +++ b/test_tipc/config/VisionTransformer/ViT_large_patch32_384_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,384,384]}] \ No newline at end of file diff --git a/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_amp_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..34fcaf290cddab226ee70b4be3caae2c10739f7b --- /dev/null +++ b/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:ViT_small_patch16_224 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_small_patch16_224.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_small_patch16_224.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/VisionTransformer/ViT_small_patch16_224.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ViT_small_patch16_224_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_infer_python.txt b/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_infer_python.txt index 6fc2cd03a44ed985da1d9a5872b7b018b9c95dbc..2098dcf1064bd4b35582e611117adab19bae2b75 100644 --- a/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_infer_python.txt +++ b/test_tipc/config/VisionTransformer/ViT_small_patch16_224_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,224,224]}] \ No newline at end of file diff --git a/test_tipc/config/Xception/Xception41_deeplab_train_amp_infer_python.txt b/test_tipc/config/Xception/Xception41_deeplab_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..325e0ad83092e10098496fc3224ff444aa291298 --- /dev/null +++ b/test_tipc/config/Xception/Xception41_deeplab_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception41_deeplab +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception41_deeplab.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception41_deeplab.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception41_deeplab.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_deeplab_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Xception/Xception41_deeplab_train_infer_python.txt b/test_tipc/config/Xception/Xception41_deeplab_train_infer_python.txt index b9c162150b8aff1a962a5ce0299cbbc6fb70e189..b4b97c325b4664b8f56cca28a9f7f47209675b1a 100644 --- a/test_tipc/config/Xception/Xception41_deeplab_train_infer_python.txt +++ b/test_tipc/config/Xception/Xception41_deeplab_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,299,299]}] \ No newline at end of file diff --git a/test_tipc/config/Xception/Xception41_train_amp_infer_python.txt b/test_tipc/config/Xception/Xception41_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..f4569e71a7d75aa1dc2a9fd09c923ac68a4aec7c --- /dev/null +++ b/test_tipc/config/Xception/Xception41_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception41 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception41.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception41.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception41.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception41_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Xception/Xception41_train_infer_python.txt b/test_tipc/config/Xception/Xception41_train_infer_python.txt index 7d7fbf10c535e54f7fd494493f9788fd924dd637..14f235d9dd70b8b279a88522ad263d0daad2a8ab 100644 --- a/test_tipc/config/Xception/Xception41_train_infer_python.txt +++ b/test_tipc/config/Xception/Xception41_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,299,299]}] \ No newline at end of file diff --git a/test_tipc/config/Xception/Xception65_deeplab_train_amp_infer_python.txt b/test_tipc/config/Xception/Xception65_deeplab_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..709423e618cf263d739c9b0eb797893b917c0f93 --- /dev/null +++ b/test_tipc/config/Xception/Xception65_deeplab_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception65_deeplab +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception65_deeplab.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception65_deeplab.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception65_deeplab.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_deeplab_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Xception/Xception65_deeplab_train_infer_python.txt b/test_tipc/config/Xception/Xception65_deeplab_train_infer_python.txt index adf12cb8f2c845660f700bf6b7d3b67db097a77e..c07b8b85082cf7145e67b59da618e957e57f1c29 100644 --- a/test_tipc/config/Xception/Xception65_deeplab_train_infer_python.txt +++ b/test_tipc/config/Xception/Xception65_deeplab_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,299,299]}] \ No newline at end of file diff --git a/test_tipc/config/Xception/Xception65_train_amp_infer_python.txt b/test_tipc/config/Xception/Xception65_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..ed1d042d3a2cee34b1e03ef834e3084d6ab2d006 --- /dev/null +++ b/test_tipc/config/Xception/Xception65_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception65 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception65.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception65.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception65.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception65_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Xception/Xception65_train_infer_python.txt b/test_tipc/config/Xception/Xception65_train_infer_python.txt index 463e20df87f642982c237f293ab372b3a372db84..3397ef169ecade06b853aeb5a66b4c29491c4606 100644 --- a/test_tipc/config/Xception/Xception65_train_infer_python.txt +++ b/test_tipc/config/Xception/Xception65_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,299,299]}] \ No newline at end of file diff --git a/test_tipc/config/Xception/Xception71_train_amp_infer_python.txt b/test_tipc/config/Xception/Xception71_train_amp_infer_python.txt new file mode 100644 index 0000000000000000000000000000000000000000..60867581f598b68c487673b71f47a94df1a318f0 --- /dev/null +++ b/test_tipc/config/Xception/Xception71_train_amp_infer_python.txt @@ -0,0 +1,52 @@ +===========================train_params=========================== +model_name:Xception71 +python:python3.7 +gpu_list:0|0,1 +-o Global.device:gpu +-o Global.auto_cast:null +-o Global.epochs:lite_train_lite_infer=2|whole_train_whole_infer=120 +-o Global.output_dir:./output/ +-o DataLoader.Train.sampler.batch_size:8 +-o Global.pretrained_model:null +train_model_name:latest +train_infer_img_dir:./dataset/ILSVRC2012/val +null:null +## +trainer:amp_train +amp_train:tools/train.py -c ppcls/configs/ImageNet/Xception/Xception71.yaml -o Global.seed=1234 -o DataLoader.Train.sampler.shuffle=False -o DataLoader.Train.loader.num_workers=0 -o DataLoader.Train.loader.use_shared_memory=False -o AMP.scale_loss=128 -o AMP.use_dynamic_loss_scaling=True -o AMP.level=O2 +pact_train:null +fpgm_train:null +distill_train:null +null:null +null:null +## +===========================eval_params=========================== +eval:tools/eval.py -c ppcls/configs/ImageNet/Xception/Xception71.yaml +null:null +## +===========================infer_params========================== +-o Global.save_inference_dir:./inference +-o Global.pretrained_model: +norm_export:tools/export_model.py -c ppcls/configs/ImageNet/Xception/Xception71.yaml +quant_export:null +fpgm_export:null +distill_export:null +kl_quant:null +export2:null +pretrained_model_url:https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/Xception71_pretrained.pdparams +infer_model:../inference/ +infer_export:True +infer_quant:Fasle +inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.transform_ops.0.ResizeImage.resize_short=320 -o PreProcess.transform_ops.1.CropImage.size=299 +-o Global.use_gpu:True|False +-o Global.enable_mkldnn:True|False +-o Global.cpu_num_threads:1|6 +-o Global.batch_size:1|16 +-o Global.use_tensorrt:True|False +-o Global.use_fp16:True|False +-o Global.inference_model_dir:../inference +-o Global.infer_imgs:../dataset/ILSVRC2012/val +-o Global.save_log_path:null +-o Global.benchmark:True +null:null +null:null diff --git a/test_tipc/config/Xception/Xception71_train_infer_python.txt b/test_tipc/config/Xception/Xception71_train_infer_python.txt index 52ea37d055e14f1ab81b65ff5e3dbc7dac8b7cb9..3a02806e48becba1b0c50995fc42943b42d25eb8 100644 --- a/test_tipc/config/Xception/Xception71_train_infer_python.txt +++ b/test_tipc/config/Xception/Xception71_train_infer_python.txt @@ -50,3 +50,5 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran -o Global.benchmark:True null:null null:null +===========================infer_benchmark_params========================== +random_infer_input:[{float32,[3,299,299]}] \ No newline at end of file diff --git a/test_tipc/docs/benchmark_train.md b/test_tipc/docs/benchmark_train.md new file mode 100644 index 0000000000000000000000000000000000000000..20cf9287423616103609187d2104d68c77f34650 --- /dev/null +++ b/test_tipc/docs/benchmark_train.md @@ -0,0 +1,53 @@ + +# TIPC Linux端Benchmark测试文档 + +该文档为Benchmark测试说明，Benchmark预测功能测试的主程序为`benchmark_train.sh`，用于验证监控模型训练的性能。 + +# 1. 测试流程 +## 1.1 准备数据和环境安装 +运行`test_tipc/prepare.sh`，完成训练数据准备和安装环境流程。 + +```shell +# 运行格式：bash test_tipc/prepare.sh train_benchmark.txt mode +bash test_tipc/prepare.sh test_tipc/configs/MobileNetV2/MobileNetV2_train_infer_python.txt benchmark_train +``` + +## 1.2 功能测试 +执行`test_tipc/benchmark_train.sh`，完成模型训练和日志解析 + +```shell +# 运行格式：bash test_tipc/benchmark_train.sh train_benchmark.txt mode +bash test_tipc/benchmark_train.sh test_tipc/config/MobileNetV2/MobileNetV2_train_infer_python.txt benchmark_train + +``` + +`test_tipc/benchmark_train.sh`支持根据传入的第三个参数实现只运行某一个训练配置，如下： +```shell +# 运行格式：bash test_tipc/benchmark_train.sh train_benchmark.txt mode params +bash test_tipc/benchmark_train.sh test_tipc/configs/MobileNetV2/MobileNetV2_train_infer_python.txt benchmark_train dynamic_bs8_fp32_DP_N1C1 +``` +dynamic_bs8_fp32_DP_N1C1为test_tipc/benchmark_train.sh传入的参数，格式如下： +`${modeltype}_${batch_size}_${fp_item}_${run_mode}_${device_num}` +包含的信息有：模型类型、batchsize大小、训练精度如fp32,fp16等、分布式运行模式以及分布式训练使用的机器信息如单机单卡（N1C1）。 + + +## 2. 日志输出 + +运行后将保存模型的训练日志和解析日志，使用 `test_tipc/configs/MobileNetV2/MobileNetV2_train_infer_python.txt` 参数文件的训练日志解析结果是： + +``` +{"model_branch": "dygaph", "model_commit": "7c39a1996b19087737c05d883fd346d2f39dbcc0", "model_name": "cls_MobileNetV2_bs8_fp32_SingleP_DP", "batch_size": 8, "fp_item": "fp32", "run_process_type": "SingleP", "run_mode": "DP", "convergence_value": "5.413110", "convergence_key": "loss:", "ips": 19.333, "speed_unit": "samples/s", "device_num": "N1C1", "model_run_time": "0", "frame_commit": "8cc09552473b842c651ead3b9848d41827a3dbab", "frame_version": "0.0.0"} +``` + +训练日志和日志解析结果保存在benchmark_log目录下，文件组织格式如下： +``` +train_log/ +├── index +│   ├── PaddleClas_cls_MobileNetV2_bs8_fp32_SingleP_DP_N1C1_speed +│   └── PaddleClas_cls_MobileNetV2_bs8_fp32_SingleP_DP_N1C4_speed +├── profiling_log +│   └── PaddleClas_cls_MobileNetV2_bs8_fp32_SingleP_DP_N1C1_profiling +└── train_log + ├── PaddleClas_cls_MobileNetV2_bs8_fp32_SingleP_DP_N1C1_log + └── PaddleClas_cls_MobileNetV2_bs8_fp32_SingleP_DP_N1C4_log +``` diff --git a/test_tipc/docs/test_lite_arm_cpu_cpp.md b/test_tipc/docs/test_lite_arm_cpu_cpp.md new file mode 100644 index 0000000000000000000000000000000000000000..eb43a21f48e828744d804de02833f5ed19fa24e8 --- /dev/null +++ b/test_tipc/docs/test_lite_arm_cpu_cpp.md @@ -0,0 +1,44 @@ +# Lite_arm_cpp_cpu 预测功能测试 + +Lite_arm_cpp_cpu 预测功能测试的主程序为`test_lite_arm_cpu_cpp.sh`，可以测试基于 Paddle-Lite 预测库的模型推理功能。 + +## 1. 测试结论汇总 + +| 模型类型 |device | batchsize | 精度类型| 线程数 | +| :----: | :----: | :----: | :----: | :----: | +| 正常模型 | arm_cpu | 1 | FP32 | 1 | + +## 2. 测试流程 +运行环境配置请参考[文档](https://github.com/PaddlePaddle/models/blob/release/2.2/tutorials/mobilenetv3_prod/Step6/deploy/lite_infer_cpp_arm_cpu/README.md) 的内容配置 TIPC Lite 的运行环境。 + +### 2.1 功能测试 +先运行 `prepare_lite_arm_cpu_cpp.sh` 准备数据和模型，然后运行 `test_lite_arm_cpu_cpp.sh` 进行测试，最终在 `./output` 目录下生成 `lite_*.log` 后缀的日志文件。 + +```shell +bash test_tipc/prepare_lite_arm_cpu_cpp.sh test_tipc/config/MobileNetV3/MobileNetV3_large_x1_0_lite_arm_cpu_cpp.txt +``` + +运行预测指令后，在`./output`文件夹下自动会保存运行日志，包括以下文件： + +```shell +test_tipc/output/ +|- results.log # 运行指令状态的日志 +|- lite_MobileNetV3_large_x1_0_runtime_device_arm_cpu_precision_FP32_batchsize_1_threads_1.log # ARM_CPU 上 FP32 状态下，线程数设置为1，测试batch_size=1条件下的预测运行日志 +...... +``` +其中results.log中包含了每条指令的运行状态，如果运行成功会输出： + +``` +Run successfully with command - adb shell 'export LD_LIBRARY_PATH=/data/local/tmp/arm_cpu/; /data/local/tmp/arm_cpu/mobilenet_v3 /data/local/tmp/arm_cpu/config.txt /data/local/tmp/arm_cpu/demo.jpg' > ./output/lite_MobileNetV3_large_x1_0_runtime_device_arm_cpu_precision_FP32_batchsize_1_threads_1.log 2>&1! +...... +``` +如果运行失败，会输出： +``` +Run failed with command - adb shell 'export LD_LIBRARY_PATH=/data/local/tmp/arm_cpu/; /data/local/tmp/arm_cpu/mobilenet_v3 /data/local/tmp/arm_cpu/config.txt /data/local/tmp/arm_cpu/demo.jpg' > ./output/lite_MobileNetV3_large_x1_0_runtime_device_arm_cpu_precision_FP32_batchsize_1_threads_1.log 2>&1! +...... +``` +可以很方便的根据results.log中的内容判定哪一个指令运行错误。 + +## 3. 更多教程 + +本文档为功能测试用，更详细的 Lite 预测使用教程请参考：[PaddleLite 推理部署](../../docs/zh_CN/inference_deployment/paddle_lite_deploy.md) 。 diff --git a/test_tipc/prepare.sh b/test_tipc/prepare.sh index acb2eb05a7e553aeae1b5c92f73e48bbfee59440..70040dc8b28656f7fb3e1384f840f068437dcf7e 100644 --- a/test_tipc/prepare.sh +++ b/test_tipc/prepare.sh @@ -1,7 +1,7 @@ #!/bin/bash FILENAME=$1 -# MODE be one of ['lite_train_lite_infer' 'lite_train_whole_infer' 'whole_train_whole_infer', +# MODE be one of ['lite_train_lite_infer' 'lite_train_whole_infer' 'whole_train_whole_infer', # 'whole_infer', 'klquant_whole_infer', # 'cpp_infer', 'serving_infer', 'lite_infer'] @@ -67,9 +67,9 @@ if [ ${MODE} = "cpp_infer" ];then model_dir=${tar_name%.*} eval "tar xf ${tar_name}" eval "mv ${model_dir} ${cls_inference_model_dir}" - + eval "wget -nc $det_inference_url" - tar_name=$(func_get_url_file_name "$det_inference_url") + tar_name=$(func_get_url_file_name "$det_inference_url") model_dir=${tar_name%.*} eval "tar xf ${tar_name}" eval "mv ${model_dir} ${det_inference_model_dir}" @@ -120,7 +120,7 @@ if [ ${MODE} = "lite_train_lite_infer" ] || [ ${MODE} = "lite_train_whole_infer" wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_little_train.tar tar xf whole_chain_little_train.tar ln -s whole_chain_little_train ILSVRC2012 - cd ILSVRC2012 + cd ILSVRC2012 mv train.txt train_list.txt mv val.txt val_list.txt cp -r train/* val/ @@ -132,7 +132,7 @@ elif [ ${MODE} = "whole_infer" ] || [ ${MODE} = "klquant_whole_infer" ];then wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_infer.tar tar xf whole_chain_infer.tar ln -s whole_chain_infer ILSVRC2012 - cd ILSVRC2012 + cd ILSVRC2012 mv val.txt val_list.txt ln -s val_list.txt train_list.txt cd ../../ @@ -153,7 +153,7 @@ elif [ ${MODE} = "whole_train_whole_infer" ];then wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/whole_chain/whole_chain_CIFAR100.tar tar xf whole_chain_CIFAR100.tar ln -s whole_chain_CIFAR100 ILSVRC2012 - cd ILSVRC2012 + cd ILSVRC2012 mv train.txt train_list.txt mv test.txt val_list.txt cd ../../ @@ -182,3 +182,15 @@ if [ ${MODE} = "paddle2onnx_infer" ];then wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/ResNet50_vd_infer.tar && tar xf ResNet50_vd_infer.tar cd ../../ fi + +if [ ${MODE} = "benchmark_train" ];then + pip install -r requirements.txt + cd dataset + rm -rf ILSVRC2012 + wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/ImageNet1k/ILSVRC2012_val.tar + tar xf ILSVRC2012_val.tar + ln -s ILSVRC2012_val ILSVRC2012 + cd ILSVRC2012 + ln -s val_list.txt train_list.txt + cd ../../ +fi diff --git a/test_tipc/prepare_lite_arm_cpu_cpp.sh b/test_tipc/prepare_lite_arm_cpu_cpp.sh new file mode 100644 index 0000000000000000000000000000000000000000..ee3e9e427992c303c84720f6a60b44a47279b850 --- /dev/null +++ b/test_tipc/prepare_lite_arm_cpu_cpp.sh @@ -0,0 +1,58 @@ +#!/bin/bash +source test_tipc/common_func.sh + +BASIC_CONFIG="./config.txt" +CONFIG=$1 + +# parser tipc config +IFS=$'\n' +TIPC_CONFIG=$1 +tipc_dataline=$(cat $TIPC_CONFIG) +tipc_lines=(${tipc_dataline}) + +runtime_device=$(func_parser_value_lite "${tipc_lines[0]}" ":") +lite_arm_work_path=$(func_parser_value_lite "${tipc_lines[1]}" ":") +lite_arm_so_path=$(func_parser_value_lite "${tipc_lines[2]}" ":") +clas_model_name=$(func_parser_value_lite "${tipc_lines[3]}" ":") +inference_cmd=$(func_parser_value_lite "${tipc_lines[4]}" ":") +num_threads_list=$(func_parser_value_lite "${tipc_lines[5]}" ":") +batch_size_list=$(func_parser_value_lite "${tipc_lines[6]}" ":") +precision_list=$(func_parser_value_lite "${tipc_lines[7]}" ":") + + +# Prepare config and test.sh +work_path="./deploy/lite" +cp ${CONFIG} ${work_path} +cp test_tipc/test_lite_arm_cpu_cpp.sh ${work_path} + +# Prepare model +cd ${work_path} +pip3 install paddlelite==2.10 +model_url="https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/${clas_model_name}_infer.tar" +wget --no-proxy ${model_url} +model_tar=$(echo ${model_url} | awk -F "/" '{print $NF}') +tar -xf ${model_tar} +paddle_lite_opt --model_dir=${clas_model_name}_infer --model_file=${clas_model_name}_infer/inference.pdmodel --param_file=${clas_model_name}_infer/inference.pdiparams --valid_targets=arm --optimize_out=${clas_model_name} +rm -rf ${clas_model_name}_infer* + +# Prepare paddlelite library +paddlelite_lib_url="https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10/inference_lite_lib.android.armv8.clang.c++_static.with_extra.with_cv.tar.gz" +wget ${paddlelite_lib_url} +paddlelite_lib_file=$(echo ${paddlelite_lib_url} | awk -F "/" '{print $NF}') +tar -xzf ${paddlelite_lib_file} +mv ${paddlelite_lib_file%*.tar.gz} inference_lite_lib.android.armv8 +rm -rf ${paddlelite_lib_file%*.tar.gz}* + +# Compile and obtain executable binary file +git clone https://github.com/LDOUBLEV/AutoLog.git +make + +# push executable binary, library, lite model, data, etc. to arm device +adb shell mkdir -p ${lite_arm_work_path} +adb push $(echo ${inference_cmd} | awk '{print $1}') ${lite_arm_work_path} +adb shell chmod +x ${lite_arm_work_path}/$(echo ${inference_cmd} | awk '{print $1}') +adb push ${lite_arm_so_path} ${lite_arm_work_path} +adb push ${clas_model_name}.nb ${lite_arm_work_path} +adb push ${BASIC_CONFIG} ${lite_arm_work_path} +adb push ../../ppcls/utils/imagenet1k_label_list.txt ${lite_arm_work_path} +adb push imgs/$(echo ${inference_cmd} | awk '{print $3}') ${lite_arm_work_path} diff --git a/test_tipc/prepare_lite_cpp.sh b/test_tipc/prepare_lite_cpp.sh deleted file mode 100644 index b129322ddde36e496c8bc1b08f7b78e94e27288f..0000000000000000000000000000000000000000 --- a/test_tipc/prepare_lite_cpp.sh +++ /dev/null @@ -1,93 +0,0 @@ -#!/bin/bash -source ./test_tipc/common_func.sh -FILENAME=$1 -dataline=$(cat ${FILENAME}) -# parser params -IFS=$'\n' -lines=(${dataline}) -IFS=$'\n' - -inference_cmd=$(func_parser_value "${lines[1]}") -DEVICE=$(func_parser_value "${lines[2]}") -det_lite_model_list=$(func_parser_value "${lines[3]}") -rec_lite_model_list=$(func_parser_value "${lines[4]}") -cls_lite_model_list=$(func_parser_value "${lines[5]}") - -if [[ $inference_cmd =~ "det" ]];then - lite_model_list=${det_lite_model_list} -elif [[ $inference_cmd =~ "rec" ]];then - lite_model_list=(${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) -elif [[ $inference_cmd =~ "system" ]];then - lite_model_list=(${det_lite_model_list[*]} ${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) -else - echo "inference_cmd is wrong, please check." - exit 1 -fi - -if [ ${DEVICE} = "ARM_CPU" ];then - valid_targets="arm" - paddlelite_url="https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10-rc/inference_lite_lib.android.armv8.gcc.c++_shared.with_extra.with_cv.tar.gz" - end_index="66" -elif [ ${DEVICE} = "ARM_GPU_OPENCL" ];then - valid_targets="opencl" - paddlelite_url="https://github.com/PaddlePaddle/Paddle-Lite/releases/download/v2.10-rc/inference_lite_lib.armv8.clang.with_exception.with_extra.with_cv.opencl.tar.gz" - end_index="71" -else - echo "DEVICE only suport ARM_CPU, ARM_GPU_OPENCL." - exit 2 -fi - -# prepare lite .nb model -pip install paddlelite==2.10-rc -current_dir=${PWD} -IFS="|" -model_path=./inference_models - -for model in ${lite_model_list[*]}; do - if [[ $model =~ "PP-OCRv2" ]];then - inference_model_url=https://paddleocr.bj.bcebos.com/PP-OCRv2/chinese/${model}.tar - elif [[ $model =~ "v2.0" ]];then - inference_model_url=https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/${model}.tar - else - echo "Model is wrong, please check." - exit 3 - fi - inference_model=${inference_model_url##*/} - wget -nc -P ${model_path} ${inference_model_url} - cd ${model_path} && tar -xf ${inference_model} && cd ../ - model_dir=${model_path}/${inference_model%.*} - model_file=${model_dir}/inference.pdmodel - param_file=${model_dir}/inference.pdiparams - paddle_lite_opt --model_dir=${model_dir} --model_file=${model_file} --param_file=${param_file} --valid_targets=${valid_targets} --optimize_out=${model_dir}_opt -done - -# prepare test data -data_url=https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015_lite.tar -model_path=./inference_models -inference_model=${inference_model_url##*/} -data_file=${data_url##*/} -wget -nc -P ./inference_models ${inference_model_url} -wget -nc -P ./test_data ${data_url} -cd ./inference_models && tar -xf ${inference_model} && cd ../ -cd ./test_data && tar -xf ${data_file} && rm ${data_file} && cd ../ - -# prepare lite env -paddlelite_zipfile=$(echo $paddlelite_url | awk -F "/" '{print $NF}') -paddlelite_file=${paddlelite_zipfile:0:${end_index}} -wget ${paddlelite_url} && tar -xf ${paddlelite_zipfile} -mkdir -p ${paddlelite_file}/demo/cxx/ocr/test_lite -cp -r ${model_path}/*_opt.nb test_data ${paddlelite_file}/demo/cxx/ocr/test_lite -cp ppocr/utils/ppocr_keys_v1.txt deploy/lite/config.txt ${paddlelite_file}/demo/cxx/ocr/test_lite -cp -r ./deploy/lite/* ${paddlelite_file}/demo/cxx/ocr/ -cp ${paddlelite_file}/cxx/lib/libpaddle_light_api_shared.so ${paddlelite_file}/demo/cxx/ocr/test_lite -cp ${FILENAME} test_tipc/test_lite_arm_cpp.sh test_tipc/common_func.sh ${paddlelite_file}/demo/cxx/ocr/test_lite -cd ${paddlelite_file}/demo/cxx/ocr/ -git clone https://github.com/cuicheng01/AutoLog.git - -# make -make -j -sleep 1 -make -j -cp ocr_db_crnn test_lite && cp test_lite/libpaddle_light_api_shared.so test_lite/libc++_shared.so -tar -cf test_lite.tar ./test_lite && cp test_lite.tar ${current_dir} && cd ${current_dir} -rm -rf ${paddlelite_file}* && rm -rf ${model_path} diff --git a/test_tipc/static/README.md b/test_tipc/static/README.md new file mode 100644 index 0000000000000000000000000000000000000000..feb980e26f6442ef818252851a2ee02a17031f45 --- /dev/null +++ b/test_tipc/static/README.md @@ -0,0 +1,36 @@ +# PaddleClas 下静态图benchmark模型执行说明 + +静态图benchmark测试脚本说明 + +# 目录说明 +# Docker 运行环境 +docker image: registry.baidubce.com/paddlepaddle/paddle:latest-dev-cuda11.2-cudnn8-gcc82 +paddle = 2.2.2 +python = 3.7 +# 运行benchmark测试步骤 + +```shell +git clone https://github.com/PaddlePaddle/PaddleClas.git +cd PaddleClas +``` + +# 准备数据 + +```shell +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +``` + +# 运行模型 +## 单卡（自动运行打开Profiling） + +```shell +export CUDA_VISIBLE_DEVICES=0 +bash test_tipc/static/${model_item}/N1C1/${shell_name}.sh +``` + +## 多卡 + +```shell +export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 +bash test_tipc/static/${model_item}/N1C8/${shell_name}.sh +``` diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..fabbb9fe62d23aea28d9b09a76bf3ce9174c3fb3 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=amp_fp16 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_fp32_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..f74de8c6293c7d9aa9a63fa349727959796ea9e5 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_fp32_DP.sh @@ -0,0 +1,16 @@ +model_item=ResNet50 +bs_item=128 +fp_item=fp32 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; +# run profiling +sleep 10; +export PROFILING=true +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..86242916f4fbf915f835bcc3f09178ed0fed5722 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs128_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=pure_fp16 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..878cd119e1b2b108d6d7329b4ee650c9f7e9c926 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=amp_fp16 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_fp32_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..0fc06ebaae4759d98cb1b8ada33ae8181dfee604 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_fp32_DP.sh @@ -0,0 +1,16 @@ +model_item=ResNet50 +bs_item=256 +fp_item=fp32 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; +# run profiling +sleep 10; +export PROFILING=true +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..47e94fb4b8df66fcf14dcd9f0a9e644f47ed4f92 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs256_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=pure_fp16 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..2d6f17c87589963bbadfdad64d5a698b9258c03f --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=amp_fp16 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_fp32_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..3db66a7e6f255c8977cb398b0750a7fd9486fc05 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_fp32_DP.sh @@ -0,0 +1,16 @@ +model_item=ResNet50 +bs_item=64 +fp_item=fp32 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; +# run profiling +sleep 10; +export PROFILING=true +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..52b8ccc5a3c18f45062d8c06b20fc8b523eec48d --- /dev/null +++ b/test_tipc/static/ResNet50/N1C1/ResNet50_bs64_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=pure_fp16 +run_mode=DP +device_num=N1C1 +max_epochs=1 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..e12494ff5e3ddaed236835cc7d6d6cbae33b69af --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=amp_fp16 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_fp32_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..e4de7513345f0066f8eabff0b682f862c02c9a55 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_fp32_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=fp32 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..08503177f89e75365ba0af8c26d69452067b9146 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs128_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=pure_fp16 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..ecabf01177abc7781ad04690f6dfdb26aa1fdc3e --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=amp_fp16 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_fp32_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..1bc077764a9f23bdc71457a8def130eb92af3264 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_fp32_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=fp32 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..0296cb7b200b959795840359bc734653471cb788 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs256_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=pure_fp16 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..ba666114be5fd7053efafc0ab8a6bcdc43513457 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=amp_fp16 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_fp32_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..8a70283991698e0b358533a598da248745c1d664 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_fp32_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=fp32 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..a7cd97ea80d30165729d19fa9f0b0a89cdf7e093 --- /dev/null +++ b/test_tipc/static/ResNet50/N1C8/ResNet50_bs64_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=pure_fp16 +run_mode=DP +device_num=N1C8 +max_epochs=8 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..c96d5603d03ebacafa02c3fb49e0c8bb6b89f531 --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=amp_fp16 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_fp32_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..f32990bd91befb1c23e39e8b9d27a50fbf5c940e --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_fp32_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=fp32 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..ef3876f1d3348c21df1de3526e162cc2ef34d8d1 --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs128_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=128 +fp_item=pure_fp16 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..8ec70d35c3aed8814e43062f70223d4a2c5fffe8 --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=amp_fp16 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_fp32_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..6ab1ec00cfc97b9b15e392bcc08ac5cde7a896e5 --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_fp32_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=fp32 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..672fb24660ba10f181098dc2c8b2cc52463bfc40 --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs256_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=256 +fp_item=pure_fp16 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_amp_fp16_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_amp_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..58f68d234d91bd98f4027d739070b045cd33a235 --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_amp_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=amp_fp16 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_fp32_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_fp32_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..eafd988120f6c836b90eb54d4c0d2bd802705773 --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_fp32_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=fp32 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_pure_fp16_DP.sh b/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_pure_fp16_DP.sh new file mode 100644 index 0000000000000000000000000000000000000000..c764a79e574bfaa5356d180321537dcb6d1fb5bb --- /dev/null +++ b/test_tipc/static/ResNet50/N4C32/ResNet50_bs64_pure_fp16_DP.sh @@ -0,0 +1,12 @@ +model_item=ResNet50 +bs_item=64 +fp_item=pure_fp16 +run_mode=DP +device_num=N4C32 +max_epochs=32 +num_workers=8 + +# get data +bash test_tipc/static/${model_item}/benchmark_common/prepare.sh +# run +bash test_tipc/static/${model_item}/benchmark_common/run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} ${max_epochs} ${num_workers} 2>&1; diff --git a/test_tipc/static/ResNet50/benchmark_common/prepare.sh b/test_tipc/static/ResNet50/benchmark_common/prepare.sh new file mode 100644 index 0000000000000000000000000000000000000000..bdf34737f889c8baedc4e1d11bca00ea9fe8a00b --- /dev/null +++ b/test_tipc/static/ResNet50/benchmark_common/prepare.sh @@ -0,0 +1,11 @@ +#!/bin/bash + +pip install -r requirements.txt +cd dataset +rm -rf ILSVRC2012 +wget -nc https://paddle-imagenet-models-name.bj.bcebos.com/data/ImageNet1k/ILSVRC2012_val.tar +tar xf ILSVRC2012_val.tar +ln -s ILSVRC2012_val ILSVRC2012 +cd ILSVRC2012 +ln -s val_list.txt train_list.txt +cd ../../ diff --git a/test_tipc/static/ResNet50/benchmark_common/run_benchmark.sh b/test_tipc/static/ResNet50/benchmark_common/run_benchmark.sh new file mode 100644 index 0000000000000000000000000000000000000000..c84ce46beb78226cf95d04f1b9f5f264882b318b --- /dev/null +++ b/test_tipc/static/ResNet50/benchmark_common/run_benchmark.sh @@ -0,0 +1,94 @@ +#!/usr/bin/env bash +# Test training benchmark for a model. +# Usage：bash run_benchmark.sh ${model_item} ${bs_item} ${fp_item} ${run_mode} ${device_num} +function _set_params(){ + model_item=${1:-"model_item"} # (必选) 模型 item + base_batch_size=${2:-"2"} # (必选) 如果是静态图单进程，则表示每张卡上的BS，需在训练时*卡数 + fp_item=${3:-"fp32"} # (必选) fp32|fp16 + run_mode=${4:-"DP"} # (必选) MP模型并行|DP数据并行|PP流水线并行|混合并行DP1-MP1-PP1|DP1-MP4-PP1 + device_num=${5:-"N1C1"} # (必选) 使用的卡数量，N1C1|N1C8|N4C32 （4机32卡） + profiling=${PROFILING:-"false"} # (必选) Profiling 开关，默认关闭，通过全局变量传递 + model_repo="PaddleClas" # (必选) 模型套件的名字 + speed_unit="samples/sec" # (必选)速度指标单位 + skip_steps=10 # (必选)解析日志，跳过模型前几个性能不稳定的step + keyword="ips:" # (必选)解析日志，筛选出性能数据所在行的关键字 + convergence_key="loss:" # (可选)解析日志，筛选出收敛数据所在行的关键字 如：convergence_key="loss:" + max_epochs=${6:-"1"} # （可选）需保证模型执行时间在5分钟内，需要修改代码提前中断的直接提PR 合入套件；或使用max_epoch参数 + num_workers=${7:-"4"} # (可选) +# 以下为通用执行命令，无特殊可不用修改 + model_name=${model_item}_bs${base_batch_size}_${fp_item}_${run_mode} # (必填) 且格式不要改动,与竞品名称对齐 + device=${CUDA_VISIBLE_DEVICES//,/ } + arr=(${device}) + num_gpu_devices=${#arr[*]} + run_log_path=${TRAIN_LOG_DIR:-$(pwd)} # （必填） TRAIN_LOG_DIR benchmark框架设置该参数为全局变量 + profiling_log_path=${PROFILING_LOG_DIR:-$(pwd)} # （必填） PROFILING_LOG_DIR benchmark框架设置该参数为全局变量 + speed_log_path=${LOG_PATH_INDEX_DIR:-$(pwd)} + + train_log_file=${run_log_path}/${model_repo}_${model_name}_${device_num}_log + profiling_log_file=${profiling_log_path}/${model_repo}_${model_name}_${device_num}_profiling + speed_log_file=${speed_log_path}/${model_repo}_${model_name}_${device_num}_speed +} +function _train(){ + batch_size=${base_batch_size} # 如果模型跑多卡单进程时,请在_train函数中计算出多卡需要的bs + echo "current CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES}, model_name=${model_name}, device_num=${device_num}, is profiling=${profiling}" + + if [ ${fp_item} = "fp32" ]; then + config_file="-c ppcls/configs/ImageNet/ResNet/ResNet50.yaml" + elif [ ${fp_item} = "amp_fp16" ]; then + config_file="-c ppcls/configs/ImageNet/ResNet/ResNet50_amp_O1_ultra.yaml" + elif [ ${fp_item} = "pure_fp16" ]; then + config_file="-c ppcls/configs/ImageNet/ResNet/ResNet50_amp_O2_ultra.yaml" + fi + if [ ${profiling} = "false" ]; then + profiling_config="" + log_file=${train_log_file} + else + profiling_config="--profiler_options=\"batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile\"" + log_file=${profiling_log_file} + fi + + train_cmd="${config_file} -o DataLoader.Train.sampler.batch_size=${base_batch_size} -o Global.epochs=${max_epochs} -o DataLoader.Train.loader.num_workers=${num_workers} ${profiling_config} -o Global.eval_during_train=False -o fuse_elewise_add_act_ops=True -o enable_addto=True" +# 以下为通用执行命令，无特殊可不用修改 + case ${run_mode} in + DP) if [[ ${device_num} = "N1C1" ]];then + echo "run ${run_mode} ${device_num}" + train_cmd="python ppcls/static/train.py ${train_cmd}" + else + rm -rf ./mylog + train_cmd="python -m paddle.distributed.launch --gpus 0,1,2,3,4,5,6,7 ppcls/static/train.py ${train_cmd}" + fi + ;; + DP1-MP1-PP1) echo "run run_mode: DP1-MP1-PP1" ;; + *) echo "choose run_mode "; exit 1; + esac + + echo "train_cmd: ${train_cmd} log_file: ${log_file}" + timeout 5m ${train_cmd} > ${log_file} 2>&1 + if [ $? -ne 0 ];then + echo -e "${model_name}, FAIL" + else + echo -e "${model_name}, SUCCESS" + fi + # kill -9 `ps -ef|grep 'python'|awk '{print $2}'` + if [ ${device_num} != "N1C1" -a -d mylog ]; then + rm ${log_file} + cp mylog/workerlog.0 ${log_file} + fi + cd ../ +} + +function _set_env(){ + export FLAGS_fraction_of_gpu_memory_to_use=0.80 + export FLAGS_cudnn_batchnorm_spatial_persistent=1 + export FLAGS_max_inplace_grad_add=8 + export FLAGS_cudnn_exhaustive_search=1 + export FLAGS_eager_delete_tensor_gb=0.0 + export FLAGS_conv_workspace_size_limit=4000 +} + + +source ${BENCHMARK_ROOT}/scripts/run_model.sh # 在该脚本中会对符合benchmark规范的log使用analysis.py 脚本进行性能数据解析;如果不联调只想要产出训练log可以注掉本行,提交时需打开 +_set_params $@ +# _train # 如果只产出训练log,不解析,可取消注释 +_set_env +_run # 该函数在run_model.sh中,执行时会调用_train; 如果不联调只产出训练log可以注掉本行,提交时需打开 diff --git a/test_tipc/test_lite_arm_cpp.sh b/test_tipc/test_lite_arm_cpp.sh deleted file mode 100644 index c071a236bb7ea35b86b32bfc3b22e87a5aabbb93..0000000000000000000000000000000000000000 --- a/test_tipc/test_lite_arm_cpp.sh +++ /dev/null @@ -1,159 +0,0 @@ -#!/bin/bash -source ./common_func.sh -export LD_LIBRARY_PATH=${PWD}:$LD_LIBRARY_PATH - -FILENAME=$1 -dataline=$(cat $FILENAME) -# parser params -IFS=$'\n' -lines=(${dataline}) - -# parser lite inference -inference_cmd=$(func_parser_value "${lines[1]}") -runtime_device=$(func_parser_value "${lines[2]}") -det_model_list=$(func_parser_value "${lines[3]}") -rec_model_list=$(func_parser_value "${lines[4]}") -cls_model_list=$(func_parser_value "${lines[5]}") -cpu_threads_list=$(func_parser_value "${lines[6]}") -det_batch_size_list=$(func_parser_value "${lines[7]}") -rec_batch_size_list=$(func_parser_value "${lines[8]}") -infer_img_dir_list=$(func_parser_value "${lines[9]}") -config_dir=$(func_parser_value "${lines[10]}") -rec_dict_dir=$(func_parser_value "${lines[11]}") -benchmark_value=$(func_parser_value "${lines[12]}") - -if [[ $inference_cmd =~ "det" ]]; then - lite_model_list=${det_lite_model_list} -elif [[ $inference_cmd =~ "rec" ]]; then - lite_model_list=(${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) -elif [[ $inference_cmd =~ "system" ]]; then - lite_model_list=(${det_lite_model_list[*]} ${rec_lite_model_list[*]} ${cls_lite_model_list[*]}) -else - echo "inference_cmd is wrong, please check." - exit 1 -fi - -LOG_PATH="./output" -mkdir -p ${LOG_PATH} -status_log="${LOG_PATH}/results.log" - - -function func_test_det(){ - IFS='|' - _script=$1 - _det_model=$2 - _log_path=$3 - _img_dir=$4 - _config=$5 - if [[ $_det_model =~ "slim" ]]; then - precision="INT8" - else - precision="FP32" - fi - - # lite inference - for num_threads in ${cpu_threads_list[*]}; do - for det_batchsize in ${det_batch_size_list[*]}; do - _save_log_path="${_log_path}/lite_${_det_model}_runtime_device_${runtime_device}_precision_${precision}_det_batchsize_${det_batchsize}_threads_${num_threads}.log" - command="${_script} ${_det_model} ${runtime_device} ${precision} ${num_threads} ${det_batchsize} ${_img_dir} ${_config} ${benchmark_value} > ${_save_log_path} 2>&1" - eval ${command} - status_check $? "${command}" "${status_log}" - done - done -} - -function func_test_rec(){ - IFS='|' - _script=$1 - _rec_model=$2 - _cls_model=$3 - _log_path=$4 - _img_dir=$5 - _config=$6 - _rec_dict_dir=$7 - - if [[ $_det_model =~ "slim" ]]; then - _precision="INT8" - else - _precision="FP32" - fi - - # lite inference - for num_threads in ${cpu_threads_list[*]}; do - for rec_batchsize in ${rec_batch_size_list[*]}; do - _save_log_path="${_log_path}/lite_${_rec_model}_${cls_model}_runtime_device_${runtime_device}_precision_${_precision}_rec_batchsize_${rec_batchsize}_threads_${num_threads}.log" - command="${_script} ${_rec_model} ${_cls_model} ${runtime_device} ${_precision} ${num_threads} ${rec_batchsize} ${_img_dir} ${_config} ${_rec_dict_dir} ${benchmark_value} > ${_save_log_path} 2>&1" - eval ${command} - status_check $? "${command}" "${status_log}" - done - done -} - -function func_test_system(){ - IFS='|' - _script=$1 - _det_model=$2 - _rec_model=$3 - _cls_model=$4 - _log_path=$5 - _img_dir=$6 - _config=$7 - _rec_dict_dir=$8 - if [[ $_det_model =~ "slim" ]]; then - _precision="INT8" - else - _precision="FP32" - fi - - # lite inference - for num_threads in ${cpu_threads_list[*]}; do - for det_batchsize in ${det_batch_size_list[*]}; do - for rec_batchsize in ${rec_batch_size_list[*]}; do - _save_log_path="${_log_path}/lite_${_det_model}_${_rec_model}_${_cls_model}_runtime_device_${runtime_device}_precision_${_precision}_det_batchsize_${det_batchsize}_rec_batchsize_${rec_batchsize}_threads_${num_threads}.log" - command="${_script} ${_det_model} ${_rec_model} ${_cls_model} ${runtime_device} ${_precision} ${num_threads} ${det_batchsize} ${_img_dir} ${_config} ${_rec_dict_dir} ${benchmark_value} > ${_save_log_path} 2>&1" - eval ${command} - status_check $? "${command}" "${status_log}" - done - done - done -} - - -echo "################### run test ###################" - -if [[ $inference_cmd =~ "det" ]]; then - IFS="|" - det_model_list=(${det_model_list[*]}) - - for i in {0..1}; do - #run lite inference - for img_dir in ${infer_img_dir_list[*]}; do - func_test_det "${inference_cmd}" "${det_model_list[i]}_opt.nb" "${LOG_PATH}" "${img_dir}" "${config_dir}" - done - done - -elif [[ $inference_cmd =~ "rec" ]]; then - IFS="|" - rec_model_list=(${rec_model_list[*]}) - cls_model_list=(${cls_model_list[*]}) - - for i in {0..1}; do - #run lite inference - for img_dir in ${infer_img_dir_list[*]}; do - func_test_rec "${inference_cmd}" "${rec_model}_opt.nb" "${cls_model_list[i]}_opt.nb" "${LOG_PATH}" "${img_dir}" "${rec_dict_dir}" "${config_dir}" - done - done - -elif [[ $inference_cmd =~ "system" ]]; then - IFS="|" - det_model_list=(${det_model_list[*]}) - rec_model_list=(${rec_model_list[*]}) - cls_model_list=(${cls_model_list[*]}) - - for i in {0..1}; do - #run lite inference - for img_dir in ${infer_img_dir_list[*]}; do - func_test_system "${inference_cmd}" "${det_model_list[i]}_opt.nb" "${rec_model_list[i]}_opt.nb" "${cls_model_list[i]}_opt.nb" "${LOG_PATH}" "${img_dir}" "${config_dir}" "${rec_dict_dir}" - done - done -fi diff --git a/test_tipc/test_lite_arm_cpu_cpp.sh b/test_tipc/test_lite_arm_cpu_cpp.sh new file mode 100644 index 0000000000000000000000000000000000000000..86c340060296019d0aef798aacd95580a438e0ff --- /dev/null +++ b/test_tipc/test_lite_arm_cpu_cpp.sh @@ -0,0 +1,95 @@ +#!/bin/bash +source test_tipc/common_func.sh +current_path=$PWD + +IFS=$'\n' + +TIPC_CONFIG=$1 +tipc_dataline=$(cat $TIPC_CONFIG) +tipc_lines=(${tipc_dataline}) + +work_path="./deploy/lite" +cd ${work_path} + +BASIC_CONFIG="config.txt" +basic_dataline=$(cat $BASIC_CONFIG) +basic_lines=(${basic_dataline}) + +# parser basic config +label_path=$(func_parser_value_lite "${basic_lines[1]}" " ") +resize_short_size=$(func_parser_value_lite "${basic_lines[2]}" " ") +crop_size=$(func_parser_value_lite "${basic_lines[3]}" " ") +visualize=$(func_parser_value_lite "${basic_lines[4]}" " ") +enable_benchmark=$(func_parser_value_lite "${basic_lines[9]}" " ") +tipc_benchmark=$(func_parser_value_lite "${basic_lines[10]}" " ") + +# parser tipc config +runtime_device=$(func_parser_value_lite "${tipc_lines[0]}" ":") +lite_arm_work_path=$(func_parser_value_lite "${tipc_lines[1]}" ":") +lite_arm_so_path=$(func_parser_value_lite "${tipc_lines[2]}" ":") +clas_model_name=$(func_parser_value_lite "${tipc_lines[3]}" ":") +inference_cmd=$(func_parser_value_lite "${tipc_lines[4]}" ":") +num_threads_list=$(func_parser_value_lite "${tipc_lines[5]}" ":") +batch_size_list=$(func_parser_value_lite "${tipc_lines[6]}" ":") +precision_list=$(func_parser_value_lite "${tipc_lines[7]}" ":") + +LOG_PATH=${current_path}"/output" +mkdir -p ${LOG_PATH} +status_log="${LOG_PATH}/results.log" + +#run Lite TIPC +function func_test_tipc(){ + IFS="|" + _basic_config=$1 + _model_name=$2 + _log_path=$3 + for num_threads in ${num_threads_list[*]}; do + if [ $(uname) = "Darwin" ]; then + sed -i " " "s/num_threads.*/num_threads ${num_threads}/" ${_basic_config} + elif [ $(expr substr $(uname -s) 1 5) = "Linux"]; then + sed -i "s/num_threads.*/num_threads ${num_threads}/" ${_basic_config} + fi + for batch_size in ${batch_size_list[*]}; do + if [ $(uname) = "Darwin" ]; then + sed -i " " "s/batch_size.*/batch_size ${batch_size}/" ${_basic_config} + elif [ $(expr substr $(uname -s) 1 5) = "Linux"]; then + sed -i "s/batch_size.*/batch_size ${batch_size}/" ${_basic_config} + fi + for precision in ${precision_list[*]}; do + if [ $(uname) = "Darwin" ]; then + sed -i " " "s/precision.*/precision ${precision}/" ${_basic_config} + elif [ $(expr substr $(uname -s) 1 5) = "Linux"]; then + sed -i "s/precision.*/precision ${precision}/" ${_basic_config} + fi + _save_log_path="${_log_path}/lite_${_model_name}_runtime_device_${runtime_device}_precision_${precision}_batchsize_${batch_size}_threads_${num_threads}.log" + real_inference_cmd=$(echo ${inference_cmd} | awk -F " " '{print path $1" "path $2" "path $3}' path="$lite_arm_work_path") + command1="adb push ${_basic_config} ${lite_arm_work_path}" + eval ${command1} + command2="adb shell 'export LD_LIBRARY_PATH=${lite_arm_work_path}; ${real_inference_cmd}' > ${_save_log_path} 2>&1" + eval ${command2} + status_check $? "${command2}" "${status_log}" + done + done + done +} + + +echo "################### run test tipc ###################" +label_map=$(echo ${label_path} | awk -F "/" '{print $NF}') + +if [ $(uname) = "Darwin" ]; then + # for Mac + sed -i " " "s/runtime_device.*/runtime_device arm_cpu/" ${BASIC_CONFIG} + escape_lite_arm_work_path=$(echo ${lite_arm_work_path//\//\\\/}) + sed -i " " "s/clas_model_file.*/clas_model_file ${escape_lite_arm_work_path}${clas_model_name}.nb/" ${BASIC_CONFIG} + sed -i " " "s/label_path.*/label_path ${escape_lite_arm_work_path}${label_map}/" ${BASIC_CONFIG} + sed -i " " "s/tipc_benchmark.*/tipc_benchmark 1/" ${BASIC_CONFIG} +elif [ $(expr substr $(uname -s) 1 5) = "Linux"]; then + # for Linux + sed -i "s/runtime_device.*/runtime_device arm_cpu/" ${BASIC_CONFIG} + escape_lite_arm_work_path=$(echo ${lite_arm_work_path//\//\\\/}) + sed -i "s/clas_model_file.*/clas_model_file ${escape_lite_arm_work_path}${clas_model_name}/" ${BASIC_CONFIG} + sed -i "s/label_path.*/label_path ${escape_lite_arm_work_path}${label_path}/" ${BASIC_CONFIG} + sed -i "s/tipc_benchmark.*/tipc_benchmark 1/" ${BASIC_CONFIG} +fi +func_test_tipc ${BASIC_CONFIG} ${clas_model_name} ${LOG_PATH} diff --git a/test_tipc/test_train_inference_python.sh b/test_tipc/test_train_inference_python.sh index 39503dd004a1f1811cd5dca1ab27adbf1bb51afa..a567ef3c6ae1e5a7429d4a5738cdb8ce5c6189fa 100644 --- a/test_tipc/test_train_inference_python.sh +++ b/test_tipc/test_train_inference_python.sh @@ -40,8 +40,8 @@ fpgm_key=$(func_parser_key "${lines[17]}") fpgm_trainer=$(func_parser_value "${lines[17]}") distill_key=$(func_parser_key "${lines[18]}") distill_trainer=$(func_parser_value "${lines[18]}") -trainer_key1=$(func_parser_key "${lines[19]}") -trainer_value1=$(func_parser_value "${lines[19]}") +to_static_key=$(func_parser_key "${lines[19]}") +to_static_trainer=$(func_parser_value "${lines[19]}") trainer_key2=$(func_parser_key "${lines[20]}") trainer_value2=$(func_parser_value "${lines[20]}") @@ -90,7 +90,11 @@ infer_value1=$(func_parser_value "${lines[50]}") if [ ! $epoch_num ]; then epoch_num=2 fi -LOG_PATH="./test_tipc/output" +if [ $MODE = 'benchmark_train' ]; then + epoch_num=1 +fi + +LOG_PATH="./test_tipc/output/${model_name}" mkdir -p ${LOG_PATH} status_log="${LOG_PATH}/results_python.log" @@ -242,9 +246,12 @@ else elif [ ${trainer} = "${distill_key}" ]; then run_train=${distill_trainer} run_export=${distill_export} - elif [ ${trainer} = ${trainer_key1} ]; then - run_train=${trainer_value1} - run_export=${export_value1} + # In case of @to_static, we re-used norm_traier, + # but append "-o Global.to_static=True" for config + # to trigger "apply_to_static" logic in 'engine.py' + elif [ ${trainer} = "${to_static_key}" ]; then + run_train="${norm_trainer} ${to_static_trainer}" + run_export=${norm_export} elif [[ ${trainer} = ${trainer_key2} ]]; then run_train=${trainer_value2} run_export=${export_value2} @@ -289,7 +296,7 @@ else fi # run train eval "unset CUDA_VISIBLE_DEVICES" - export FLAGS_cudnn_deterministic=True + # export FLAGS_cudnn_deterministic=True sleep 5 eval $cmd status_check $? "${cmd}" "${status_log}"