From acaf10684d8e071e48098c660355d54981d93b7c Mon Sep 17 00:00:00 2001 From: Guanghua Yu <742925032@qq.com> Date: Sat, 7 May 2022 13:37:59 +0800 Subject: [PATCH] add pp-yoloe quantization model (#5894) --- configs/slim/README.md | 3 +++ configs/slim/README_en.md | 4 +++- configs/slim/quant/ppyoloe_l_qat.yml | 26 ++++++++++++++++++++++++++ ppdet/engine/trainer.py | 7 ++++--- ppdet/modeling/heads/ppyoloe_head.py | 1 + ppdet/slim/quant.py | 5 +++++ 6 files changed, 42 insertions(+), 4 deletions(-) create mode 100644 configs/slim/quant/ppyoloe_l_qat.yml diff --git a/configs/slim/README.md b/configs/slim/README.md index ac025688b..caba1e8b2 100755 --- a/configs/slim/README.md +++ b/configs/slim/README.md @@ -126,6 +126,8 @@ python tools/export_model.py -c configs/{MODEL.yml} --slim_config configs/slim/{ | 模型 | 压缩策略 | 输入尺寸 | 模型体积(MB) | 预测时延(V100) | 预测时延(SD855) | Box AP | 下载 | Inference模型下载 | 模型配置文件 | 压缩算法配置文件 | | ------------------ | ------------ | -------- | :---------: | :---------: |:---------: | :---------: | :----------------------------------------------: | :----------------------------------------------: |:------------------------------------------: | :------------------------------------: | +| PP-YOLOE-l | baseline | 640 | - | 11.2ms(trt_fp32) | 7.7ms(trt_fp16) | -- | 50.9 | [下载链接](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | - | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | - | +| PP-YOLOE-l | 普通在线量化 | 640 | - | 6.7ms(trt_int8) | -- | 48.8 | [下载链接](https://paddledet.bj.bcebos.com/models/slim/ppyoloe_l_coco_qat.pdparams) | - | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/slim/quant/ppyoloe_l_qat.yml) | | PP-YOLOv2_R50vd | baseline | 640 | 208.6 | 19.1ms | -- | 49.1 | [下载链接](https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams) | [下载链接](https://paddledet.bj.bcebos.com/models/slim/ppyolov2_r50vd_dcn_365e_coco.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | - | | PP-YOLOv2_R50vd | PACT在线量化 | 640 | -- | 17.3ms | -- | 48.1 | [下载链接](https://paddledet.bj.bcebos.com/models/slim/ppyolov2_r50vd_dcn_qat.pdparams) | [下载链接](https://paddledet.bj.bcebos.com/models/slim/ppyolov2_r50vd_dcn_qat.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/slim/quant/ppyolov2_r50vd_dcn_qat.yml) | | PP-YOLO_R50vd | baseline | 608 | 183.3 | 17.4ms | -- | 44.8 | [下载链接](https://paddledet.bj.bcebos.com/models/ppyolo_r50vd_dcn_1x_coco.pdparams) | [下载链接](https://paddledet.bj.bcebos.com/models/slim/ppyolo_r50vd_dcn_1x_coco.tar) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | - | @@ -146,6 +148,7 @@ python tools/export_model.py -c configs/{MODEL.yml} --slim_config configs/slim/{ 说明： - 上述V100预测时延非量化模型均是使用TensorRT-FP32测试，量化模型均使用TensorRT-INT8测试，并且都包含NMS耗时。 - SD855预测时延为使用PaddleLite部署，使用arm8架构并使用4线程(4 Threads)推理时延。 +- 上述PP-YOLOE模型均在V100，开启TensorRT环境中测速，不包含NMS。（导出模型时指定：-o trt=True exclude_nms=True） ### 离线量化 需要准备val集，用来对离线量化模型进行校准，运行方式： diff --git a/configs/slim/README_en.md b/configs/slim/README_en.md index 8d2b39c91..924757e3f 100755 --- a/configs/slim/README_en.md +++ b/configs/slim/README_en.md @@ -124,7 +124,9 @@ Description: | Model | Compression Strategy | Input Size | Model Volume(MB) | Prediction Delay(V100) | Prediction Delay(SD855) | Box AP | Download | Download of Inference Model | Model Configuration File | Compression Algorithm Configuration File | | ------------------------- | -------------------------- | ----------- | :--------------: | :--------------------: | :---------------------: | :-------------------: | :-----------------------------------------------------------------------------------------: | :-----------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------: | -| PP-YOLOv2_R50vd | baseline | 640 | 208.6 | 19.1ms | -- | 49.1 | [link](https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams) | [link](https://paddledet.bj.bcebos.com/models/slim/ppyolov2_r50vd_dcn_365e_coco.tar) | [Configuration File ](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | - | +| PP-YOLOE-l | baseline | 640 | - | 11.2ms(trt_fp32) | 7.7ms(trt_fp16) | -- | 50.9 | [link](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | - | [Configuration File](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | - | +| PP-YOLOE-l | Common Online quantitative | 640 | - | 6.7ms(trt_int8) | -- | 48.8 | [link](https://paddledet.bj.bcebos.com/models/slim/ppyoloe_l_coco_qat.pdparams) | - | [Configuration File](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe/ppyoloe_crn_l_300e_coco.yml) | [Configuration File](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/slim/quant/ppyoloe_l_qat.yml) | +| PP-YOLOv2_R50vd | baseline | 640 | 208.6 | 19.1ms | -- | 49.1 | [link](https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams) | [link](https://paddledet.bj.bcebos.com/models/slim/ppyolov2_r50vd_dcn_365e_coco.tar) | [Configuration File](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | - | | PP-YOLOv2_R50vd | PACT Online quantitative | 640 | -- | 17.3ms | -- | 48.1 | [link](https://paddledet.bj.bcebos.com/models/slim/ppyolov2_r50vd_dcn_qat.pdparams) | [link](https://paddledet.bj.bcebos.com/models/slim/ppyolov2_r50vd_dcn_qat.tar) | [Configuration File ](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | [Configuration File ](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/slim/quant/ppyolov2_r50vd_dcn_qat.yml) | | PP-YOLO_R50vd | baseline | 608 | 183.3 | 17.4ms | -- | 44.8 | [link](https://paddledet.bj.bcebos.com/models/ppyolo_r50vd_dcn_1x_coco.pdparams) | [link](https://paddledet.bj.bcebos.com/models/slim/ppyolo_r50vd_dcn_1x_coco.tar) | [Configuration File ](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | - | | PP-YOLO_R50vd | PACT Online quantitative | 608 | 67.3 | 13.8ms | -- | 44.3 | [link](https://paddledet.bj.bcebos.com/models/slim/ppyolo_r50vd_qat_pact.pdparams) | [link](https://paddledet.bj.bcebos.com/models/slim/ppyolo_r50vd_qat_pact.tar) | [Configuration File ](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml) | [Configuration File ](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/slim/quant/ppyolo_r50vd_qat_pact.yml) | diff --git a/configs/slim/quant/ppyoloe_l_qat.yml b/configs/slim/quant/ppyoloe_l_qat.yml new file mode 100644 index 000000000..4c0e94003 --- /dev/null +++ b/configs/slim/quant/ppyoloe_l_qat.yml @@ -0,0 +1,26 @@ +pretrain_weights: https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams +slim: QAT + +QAT: + quant_config: { + 'weight_quantize_type': 'channel_wise_abs_max', 'activation_quantize_type': 'moving_average_abs_max', + 'weight_bits': 8, 'activation_bits': 8, 'dtype': 'int8', 'window_size': 10000, 'moving_rate': 0.9, + 'quantizable_layer_type': ['Conv2D', 'Linear']} + print_model: True + +epoch: 30 +snapshot_epoch: 5 +LearningRate: + base_lr: 0.001 + schedulers: + - !PiecewiseDecay + gamma: 0.1 + milestones: + - 10 + - 20 + - !LinearWarmup + start_factor: 0. + steps: 100 + +TrainReader: + batch_size: 8 diff --git a/ppdet/engine/trainer.py b/ppdet/engine/trainer.py index b08eb50de..b7737b1e1 100644 --- a/ppdet/engine/trainer.py +++ b/ppdet/engine/trainer.py @@ -701,9 +701,10 @@ class Trainer(object): if hasattr(self.model, 'deploy'): self.model.deploy = True - for layer in self.model.sublayers(): - if hasattr(layer, 'convert_to_deploy'): - layer.convert_to_deploy() + if 'slim' not in self.cfg: + for layer in self.model.sublayers(): + if hasattr(layer, 'convert_to_deploy'): + layer.convert_to_deploy() export_post_process = self.cfg['export'].get( 'post_process', False) if hasattr(self.cfg, 'export') else True diff --git a/ppdet/modeling/heads/ppyoloe_head.py b/ppdet/modeling/heads/ppyoloe_head.py index 70d93f2e1..3babb2703 100644 --- a/ppdet/modeling/heads/ppyoloe_head.py +++ b/ppdet/modeling/heads/ppyoloe_head.py @@ -111,6 +111,7 @@ class PPYOLOEHead(nn.Layer): in_c, 4 * (self.reg_max + 1), 3, padding=1)) # projection conv self.proj_conv = nn.Conv2D(self.reg_max + 1, 1, 1, bias_attr=False) + self.proj_conv.skip_quant = True self._init_weights() @classmethod diff --git a/ppdet/slim/quant.py b/ppdet/slim/quant.py index ab81127ae..44508198c 100644 --- a/ppdet/slim/quant.py +++ b/ppdet/slim/quant.py @@ -38,6 +38,11 @@ class QAT(object): logger.info("Model before quant:") logger.info(model) + # For PP-YOLOE, convert model to deploy firstly. + for layer in model.sublayers(): + if hasattr(layer, 'convert_to_deploy'): + layer.convert_to_deploy() + self.quanter.quantize(model) if self.print_model: -- GitLab