diff --git a/deploy/cpp_infer/src/ocr_det.cpp b/deploy/cpp_infer/src/ocr_det.cpp index 56fbace8cc6fa27f8172bed248573f15d0c98dac..bf94abce236853410c15434d494058be03a62a81 100644 --- a/deploy/cpp_infer/src/ocr_det.cpp +++ b/deploy/cpp_infer/src/ocr_det.cpp @@ -26,6 +26,8 @@ void DBDetector::LoadModel(const std::string &model_dir) { config.DisableGpu(); if (this->use_mkldnn_) { config.EnableMKLDNN(); + // cache 10 different shapes for mkldnn to avoid memory leak + config.SetMkldnnCacheCapacity(10); } config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_); } diff --git a/deploy/cpp_infer/src/ocr_rec.cpp b/deploy/cpp_infer/src/ocr_rec.cpp index e37994b562cc4bf593332432a990afe4c6697531..7f88adc54636b4ecc61d257b7cb9159ebcdb82af 100644 --- a/deploy/cpp_infer/src/ocr_rec.cpp +++ b/deploy/cpp_infer/src/ocr_rec.cpp @@ -129,6 +129,8 @@ void CRNNRecognizer::LoadModel(const std::string &model_dir) { config.DisableGpu(); if (this->use_mkldnn_) { config.EnableMKLDNN(); + // cache 10 different shapes for mkldnn to avoid memory leak + config.SetMkldnnCacheCapacity(10); } config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_); } diff --git a/deploy/cpp_infer/tools/config.txt b/deploy/cpp_infer/tools/config.txt index 17b507791cefbf6c6c15974ba92bfc0ac6b60e09..28bacba60d4a599ad951c9820938b38e55b07283 100644 --- a/deploy/cpp_infer/tools/config.txt +++ b/deploy/cpp_infer/tools/config.txt @@ -3,24 +3,25 @@ use_gpu 0 gpu_id 0 gpu_mem 4000 cpu_math_library_num_threads 10 -use_mkldnn 0 -use_zero_copy_run 0 +use_mkldnn 1 +use_zero_copy_run 1 # det config max_side_len 960 det_db_thresh 0.3 det_db_box_thresh 0.5 det_db_unclip_ratio 2.0 -det_model_dir ../model/det +det_model_dir ./inference/det_db # cls config use_angle_cls 0 -cls_model_dir ../model/cls +cls_model_dir ../inference/cls cls_thresh 0.9 # rec config -rec_model_dir ../model/rec -char_list_file ../model/ppocr_keys_v1.txt +rec_model_dir ./inference/rec_crnn +char_list_file ../../ppocr/utils/ppocr_keys_v1.txt # show the detection results -visualize 1 \ No newline at end of file +visualize 1 + diff --git a/docker/hubserving/README.md b/deploy/docker/hubserving/README.md similarity index 99% rename from docker/hubserving/README.md rename to deploy/docker/hubserving/README.md index 71e2377dcc4f7524384752b95c53f02471353f34..62381073d4c7448f9a238ca4dda4b294ce864f7a 100644 --- a/docker/hubserving/README.md +++ b/deploy/docker/hubserving/README.md @@ -20,7 +20,7 @@ git clone https://github.com/PaddlePaddle/PaddleOCR.git ``` b. Goto Dockerfile directory（ps：Need to distinguish between cpu and gpu version, the following takes cpu as an example, gpu version needs to replace the keyword） ``` -cd docker/cpu +cd deploy/docker/cpu ``` c. Build image ``` diff --git a/docker/hubserving/README_cn.md b/deploy/docker/hubserving/README_cn.md similarity index 99% rename from docker/hubserving/README_cn.md rename to deploy/docker/hubserving/README_cn.md index 9b9e5f50f5b22f3a2125a656112a20542010ac68..f117a0ab4186fea0cb94881c65b2b353bee37ff7 100644 --- a/docker/hubserving/README_cn.md +++ b/deploy/docker/hubserving/README_cn.md @@ -20,7 +20,7 @@ git clone https://github.com/PaddlePaddle/PaddleOCR.git ``` b.切换至Dockerfile目录（注：需要区分cpu或gpu版本，下文以cpu为例，gpu版本需要替换一下关键字即可） ``` -cd docker/cpu +cd deploy/docker/cpu ``` c.生成镜像 ``` diff --git a/docker/hubserving/cpu/Dockerfile b/deploy/docker/hubserving/cpu/Dockerfile similarity index 100% rename from docker/hubserving/cpu/Dockerfile rename to deploy/docker/hubserving/cpu/Dockerfile diff --git a/docker/hubserving/gpu/Dockerfile b/deploy/docker/hubserving/gpu/Dockerfile similarity index 100% rename from docker/hubserving/gpu/Dockerfile rename to deploy/docker/hubserving/gpu/Dockerfile diff --git a/docker/hubserving/sample_request.txt b/deploy/docker/hubserving/sample_request.txt similarity index 100% rename from docker/hubserving/sample_request.txt rename to deploy/docker/hubserving/sample_request.txt diff --git a/ppocr/utils/utility.py b/ppocr/utils/utility.py index e27dd1d8738a25c6a6669b99ad2b6eed4a9f25d0..2cf3c8f5c9ebba07ee1c21fe2248fe3f600126d9 100755 --- a/ppocr/utils/utility.py +++ b/ppocr/utils/utility.py @@ -90,15 +90,3 @@ def check_and_read_gif(img_path): return imgvalue, True return None, False - -def create_multi_devices_program(program, loss_var_name): - build_strategy = fluid.BuildStrategy() - build_strategy.memory_optimize = False - build_strategy.enable_inplace = True - exec_strategy = fluid.ExecutionStrategy() - exec_strategy.num_iteration_per_drop_scope = 1 - compile_program = fluid.CompiledProgram(program).with_data_parallel( - loss_name=loss_var_name, - build_strategy=build_strategy, - exec_strategy=exec_strategy) - return compile_program diff --git a/tools/infer/predict_system.py b/tools/infer/predict_system.py index bb97c8fcf4ec936309f967ca208e59876b051f17..3e6be234c68dcd82f0f9e844f3ad2859000cec88 100755 --- a/tools/infer/predict_system.py +++ b/tools/infer/predict_system.py @@ -133,7 +133,6 @@ def main(args): image_file_list = get_image_file_list(args.image_dir) text_sys = TextSystem(args) is_visualize = True - tackle_img_num = 0 for image_file in image_file_list: img, flag = check_and_read_gif(image_file) if not flag: @@ -142,9 +141,6 @@ def main(args): logger.info("error in loading image:{}".format(image_file)) continue starttime = time.time() - tackle_img_num += 1 - if not args.use_gpu and args.enable_mkldnn and tackle_img_num % 30 == 0: - text_sys = TextSystem(args) dt_boxes, rec_res = text_sys(img) elapse = time.time() - starttime print("Predict time of %s: %.3fs" % (image_file, elapse)) diff --git a/tools/infer/utility.py b/tools/infer/utility.py index ac04c2bd2aeb0e48d1db2bbf83de96af6863485d..92212afd5f3e16601939d0ca7882fb3b90c3a9ac 100755 --- a/tools/infer/utility.py +++ b/tools/infer/utility.py @@ -112,6 +112,8 @@ def create_predictor(args, mode): config.disable_gpu() config.set_cpu_math_library_num_threads(6) if args.enable_mkldnn: + # cache 10 different shapes for mkldnn to avoid memory leak + config.set_mkldnn_cache_capacity(10) config.enable_mkldnn() # config.enable_memory_optim()