PP-OCR模型量化导出模型时报错

ChihoJack commented 1 year ago

系统环境/System Environment：Windows 10 家庭版 64位
版本号/Version：Paddle：paddlepaddle-gpu 2.4.2.post112 PaddleOCR：2.6.1.3 问题相关组件/Related components：模型量化
运行指令/Command Code： python deploy/slim/quantization/export_model.py -c configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml -o Global.checkpoints=output/quant_model_distill/best_accuracy Global.save_inferenc e_dir=./output/quant_inference_model
完整报错/Complete Error Message：
[03-15 09:09:52 MainThread @logger.py:242] Argv: deploy/slim/quantization/export_model.py -c configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml -o Global.checkpoints=output/quant_model_distill/best_accuracy Global.save_inference_dir=./out put/quant_inference_model [03-15 09:09:52 MainThread @utils.py:73] paddlepaddle version: 2.4.2. 2023-03-15 09:09:53,041-WARNING: post-quant-hpo is not support in system other than linux [2023/03/15 09:09:53] ppocr INFO: Architecture : [2023/03/15 09:09:53] ppocr INFO: Models : [2023/03/15 09:09:53] ppocr INFO: Student : [2023/03/15 09:09:53] ppocr INFO: Backbone : [2023/03/15 09:09:53] ppocr INFO: disable_se : True [2023/03/15 09:09:53] ppocr INFO: model_name : small [2023/03/15 09:09:53] ppocr INFO: name : MobileNetV3 [2023/03/15 09:09:53] ppocr INFO: scale : 0.5 [2023/03/15 09:09:53] ppocr INFO: Head : [2023/03/15 09:09:53] ppocr INFO: k : 50 [2023/03/15 09:09:53] ppocr INFO: name : DBHead [2023/03/15 09:09:53] ppocr INFO: Neck : [2023/03/15 09:09:53] ppocr INFO: name : RSEFPN [2023/03/15 09:09:53] ppocr INFO: out_channels : 96 [2023/03/15 09:09:53] ppocr INFO: shortcut : True [2023/03/15 09:09:53] ppocr INFO: Transform : None [2023/03/15 09:09:53] ppocr INFO: algorithm : DB [2023/03/15 09:09:53] ppocr INFO: model_type : det [2023/03/15 09:09:53] ppocr INFO: pretrained : None [2023/03/15 09:09:53] ppocr INFO: Student2 : [2023/03/15 09:09:53] ppocr INFO: Backbone : [2023/03/15 09:09:53] ppocr INFO: disable_se : True [2023/03/15 09:09:53] ppocr INFO: model_name : small [2023/03/15 09:09:53] ppocr INFO: name : MobileNetV3 [2023/03/15 09:09:53] ppocr INFO: scale : 0.5 [2023/03/15 09:09:53] ppocr INFO: Head : [2023/03/15 09:09:53] ppocr INFO: k : 50 [2023/03/15 09:09:53] ppocr INFO: name : DBHead [2023/03/15 09:09:53] ppocr INFO: Neck : [2023/03/15 09:09:53] ppocr INFO: name : RSEFPN [2023/03/15 09:09:53] ppocr INFO: out_channels : 96 [2023/03/15 09:09:53] ppocr INFO: shortcut : True [2023/03/15 09:09:53] ppocr INFO: Transform : None [2023/03/15 09:09:53] ppocr INFO: algorithm : DB [2023/03/15 09:09:53] ppocr INFO: model_type : det [2023/03/15 09:09:53] ppocr INFO: pretrained : None [2023/03/15 09:09:53] ppocr INFO: Teacher : [2023/03/15 09:09:53] ppocr INFO: Backbone : [2023/03/15 09:09:53] ppocr INFO: in_channels : 3 [2023/03/15 09:09:53] ppocr INFO: layers : 50 [2023/03/15 09:09:53] ppocr INFO: name : ResNet_vd [2023/03/15 09:09:53] ppocr INFO: Head : [2023/03/15 09:09:53] ppocr INFO: k : 50 [2023/03/15 09:09:53] ppocr INFO: kernel_list : [7, 2, 2] [2023/03/15 09:09:53] ppocr INFO: name : DBHead [2023/03/15 09:09:53] ppocr INFO: Neck : [2023/03/15 09:09:53] ppocr INFO: name : LKPAN [2023/03/15 09:09:53] ppocr INFO: out_channels : 256 [2023/03/15 09:09:53] ppocr INFO: algorithm : DB [2023/03/15 09:09:53] ppocr INFO: freeze_params : True [2023/03/15 09:09:53] ppocr INFO: model_type : det [2023/03/15 09:09:53] ppocr INFO: return_all_feats : False [2023/03/15 09:09:53] ppocr INFO: algorithm : Distillation [2023/03/15 09:09:53] ppocr INFO: model_type : det [2023/03/15 09:09:53] ppocr INFO: name : DistillationModel [2023/03/15 09:09:53] ppocr INFO: Eval : [2023/03/15 09:09:53] ppocr INFO: dataset : [2023/03/15 09:09:53] ppocr INFO: data_dir : ./train_data/icdar2015/text_localization/ [2023/03/15 09:09:53] ppocr INFO: label_file_list : ['./train_data/icdar2015/text_localization/test_icdar2015_label.txt'] [2023/03/15 09:09:53] ppocr INFO: name : SimpleDataSet [2023/03/15 09:09:53] ppocr INFO: transforms : [2023/03/15 09:09:53] ppocr INFO: DecodeImage : [2023/03/15 09:09:53] ppocr INFO: channel_first : False [2023/03/15 09:09:53] ppocr INFO: img_mode : BGR [2023/03/15 09:09:53] ppocr INFO: DetLabelEncode : None [2023/03/15 09:09:53] ppocr INFO: DetResizeForTest : None [2023/03/15 09:09:53] ppocr INFO: NormalizeImage : [2023/03/15 09:09:53] ppocr INFO: mean : [0.485, 0.456, 0.406] [2023/03/15 09:09:53] ppocr INFO: order : hwc [2023/03/15 09:09:53] ppocr INFO: scale : 1./255. [2023/03/15 09:09:53] ppocr INFO: std : [0.229, 0.224, 0.225] [2023/03/15 09:09:53] ppocr INFO: ToCHWImage : None [2023/03/15 09:09:53] ppocr INFO: KeepKeys : [2023/03/15 09:09:53] ppocr INFO: keep_keys : ['image', 'shape', 'polys', 'ignore_tags'] [2023/03/15 09:09:53] ppocr INFO: loader : [2023/03/15 09:09:53] ppocr INFO: batch_size_per_card : 1 [2023/03/15 09:09:53] ppocr INFO: drop_last : False [2023/03/15 09:09:53] ppocr INFO: num_workers : 1 [2023/03/15 09:09:53] ppocr INFO: shuffle : False [2023/03/15 09:09:53] ppocr INFO: Global : [2023/03/15 09:09:53] ppocr INFO: cal_metric_during_train : False [2023/03/15 09:09:53] ppocr INFO: checkpoints : output/quant_model_distill/best_accuracy [2023/03/15 09:09:53] ppocr INFO: debug : False [2023/03/15 09:09:53] ppocr INFO: distributed : False [2023/03/15 09:09:53] ppocr INFO: epoch_num : 100 [2023/03/15 09:09:53] ppocr INFO: eval_batch_step : [0, 400] [2023/03/15 09:09:53] ppocr INFO: infer_img : doc/imgs_en/img_10.jpg [2023/03/15 09:09:53] ppocr INFO: log_smooth_window : 20 [2023/03/15 09:09:53] ppocr INFO: pretrained_model : None [2023/03/15 09:09:53] ppocr INFO: print_batch_step : 10 [2023/03/15 09:09:53] ppocr INFO: save_epoch_step : 100 [2023/03/15 09:09:53] ppocr INFO: save_inference_dir : ./output/quant_inference_model [2023/03/15 09:09:53] ppocr INFO: save_model_dir : ./output/ch_PP-OCR_v3_det/ [2023/03/15 09:09:53] ppocr INFO: save_res_path : ./checkpoints/det_db/predicts_db.txt [2023/03/15 09:09:53] ppocr INFO: use_gpu : True [2023/03/15 09:09:53] ppocr INFO: use_visualdl : False [2023/03/15 09:09:53] ppocr INFO: Loss : [2023/03/15 09:09:53] ppocr INFO: loss_config_list : [2023/03/15 09:09:53] ppocr INFO: DistillationDilaDBLoss : [2023/03/15 09:09:53] ppocr INFO: alpha : 5 [2023/03/15 09:09:53] ppocr INFO: balance_loss : True [2023/03/15 09:09:53] ppocr INFO: beta : 10 [2023/03/15 09:09:53] ppocr INFO: key : maps [2023/03/15 09:09:53] ppocr INFO: main_loss_type : DiceLoss [2023/03/15 09:09:53] ppocr INFO: model_name_pairs : [['Student', 'Teacher'], ['Student2', 'Teacher']] [2023/03/15 09:09:53] ppocr INFO: ohem_ratio : 3 [2023/03/15 09:09:53] ppocr INFO: weight : 1.0 [2023/03/15 09:09:53] ppocr INFO: DistillationDMLLoss : [2023/03/15 09:09:53] ppocr INFO: key : maps [2023/03/15 09:09:53] ppocr INFO: maps_name : thrink_maps [2023/03/15 09:09:53] ppocr INFO: model_name_pairs : ['Student', 'Student2'] [2023/03/15 09:09:53] ppocr INFO: weight : 1.0 [2023/03/15 09:09:53] ppocr INFO: DistillationDBLoss : [2023/03/15 09:09:53] ppocr INFO: alpha : 5 [2023/03/15 09:09:53] ppocr INFO: balance_loss : True [2023/03/15 09:09:53] ppocr INFO: beta : 10 [2023/03/15 09:09:53] ppocr INFO: main_loss_type : DiceLoss [2023/03/15 09:09:53] ppocr INFO: model_name_list : ['Student', 'Student2'] [2023/03/15 09:09:53] ppocr INFO: ohem_ratio : 3 [2023/03/15 09:09:53] ppocr INFO: weight : 1.0 [2023/03/15 09:09:53] ppocr INFO: name : CombinedLoss [2023/03/15 09:09:53] ppocr INFO: Metric : [2023/03/15 09:09:53] ppocr INFO: base_metric_name : DetMetric [2023/03/15 09:09:53] ppocr INFO: key : Student [2023/03/15 09:09:53] ppocr INFO: main_indicator : hmean [2023/03/15 09:09:53] ppocr INFO: name : DistillationMetric [2023/03/15 09:09:53] ppocr INFO: Optimizer : [2023/03/15 09:09:53] ppocr INFO: beta1 : 0.9 [2023/03/15 09:09:53] ppocr INFO: beta2 : 0.999 [2023/03/15 09:09:53] ppocr INFO: lr : [2023/03/15 09:09:53] ppocr INFO: learning_rate : 0.001 [2023/03/15 09:09:53] ppocr INFO: name : Cosine [2023/03/15 09:09:53] ppocr INFO: warmup_epoch : 2 [2023/03/15 09:09:53] ppocr INFO: name : Adam [2023/03/15 09:09:53] ppocr INFO: regularizer : [2023/03/15 09:09:53] ppocr INFO: factor : 5e-05 [2023/03/15 09:09:53] ppocr INFO: name : L2 [2023/03/15 09:09:53] ppocr INFO: PostProcess : [2023/03/15 09:09:53] ppocr INFO: box_thresh : 0.6 [2023/03/15 09:09:53] ppocr INFO: key : head_out [2023/03/15 09:09:53] ppocr INFO: max_candidates : 1000 [2023/03/15 09:09:53] ppocr INFO: model_name : ['Student'] [2023/03/15 09:09:53] ppocr INFO: name : DistillationDBPostProcess [2023/03/15 09:09:53] ppocr INFO: thresh : 0.3 [2023/03/15 09:09:53] ppocr INFO: unclip_ratio : 1.5 [2023/03/15 09:09:53] ppocr INFO: Train : [2023/03/15 09:09:53] ppocr INFO: dataset : [2023/03/15 09:09:53] ppocr INFO: data_dir : ./train_data/icdar2015/text_localization/ [2023/03/15 09:09:53] ppocr INFO: label_file_list : ['./train_data/icdar2015/text_localization/train_icdar2015_label.txt'] [2023/03/15 09:09:53] ppocr INFO: name : SimpleDataSet [2023/03/15 09:09:53] ppocr INFO: ratio_list : [1.0] [2023/03/15 09:09:53] ppocr INFO: transforms : [2023/03/15 09:09:53] ppocr INFO: DecodeImage : [2023/03/15 09:09:53] ppocr INFO: channel_first : False [2023/03/15 09:09:53] ppocr INFO: img_mode : BGR [2023/03/15 09:09:53] ppocr INFO: DetLabelEncode : None [2023/03/15 09:09:53] ppocr INFO: CopyPaste : None [2023/03/15 09:09:53] ppocr INFO: IaaAugment : [2023/03/15 09:09:53] ppocr INFO: augmenter_args : [2023/03/15 09:09:53] ppocr INFO: args : [2023/03/15 09:09:53] ppocr INFO: p : 0.5 [2023/03/15 09:09:53] ppocr INFO: type : Fliplr [2023/03/15 09:09:53] ppocr INFO: args : [2023/03/15 09:09:53] ppocr INFO: rotate : [-10, 10] [2023/03/15 09:09:53] ppocr INFO: type : Affine [2023/03/15 09:09:53] ppocr INFO: args : [2023/03/15 09:09:53] ppocr INFO: size : [0.5, 3] [2023/03/15 09:09:53] ppocr INFO: type : Resize [2023/03/15 09:09:53] ppocr INFO: EastRandomCropData : [2023/03/15 09:09:53] ppocr INFO: keep_ratio : True [2023/03/15 09:09:53] ppocr INFO: max_tries : 50 [2023/03/15 09:09:53] ppocr INFO: size : [960, 960] [2023/03/15 09:09:53] ppocr INFO: MakeBorderMap : [2023/03/15 09:09:53] ppocr INFO: shrink_ratio : 0.4 [2023/03/15 09:09:53] ppocr INFO: thresh_max : 0.7 [2023/03/15 09:09:53] ppocr INFO: thresh_min : 0.3 [2023/03/15 09:09:53] ppocr INFO: MakeShrinkMap : [2023/03/15 09:09:53] ppocr INFO: min_text_size : 8 [2023/03/15 09:09:53] ppocr INFO: shrink_ratio : 0.4 [2023/03/15 09:09:53] ppocr INFO: NormalizeImage : [2023/03/15 09:09:53] ppocr INFO: mean : [0.485, 0.456, 0.406] [2023/03/15 09:09:53] ppocr INFO: order : hwc [2023/03/15 09:09:53] ppocr INFO: scale : 1./255. [2023/03/15 09:09:53] ppocr INFO: std : [0.229, 0.224, 0.225] [2023/03/15 09:09:53] ppocr INFO: ToCHWImage : None [2023/03/15 09:09:53] ppocr INFO: KeepKeys : [2023/03/15 09:09:53] ppocr INFO: keep_keys : ['image', 'threshold_map', 'threshold_mask', 'shrink_map', 'shrink_mask'] [2023/03/15 09:09:53] ppocr INFO: loader : [2023/03/15 09:09:53] ppocr INFO: batch_size_per_card : 2 [2023/03/15 09:09:53] ppocr INFO: drop_last : False [2023/03/15 09:09:53] ppocr INFO: num_workers : 8 [2023/03/15 09:09:53] ppocr INFO: shuffle : True [2023/03/15 09:09:53] ppocr INFO: profiler_options : None [2023/03/15 09:09:53] ppocr INFO: train with paddle 2.4.2 and device Place(gpu:0) W0315 09:09:53.222466 14088 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 7.5, Driver API Version: 11.2, Runtime API Version: 11.2 W0315 09:09:53.228482 14088 gpu_resources.cc:91] device: 0, cuDNN Version: 8.8. [2023/03/15 09:09:57] ppocr INFO: resume from output/quant_model_distill/best_accuracy [2023/03/15 09:09:57] ppocr INFO: Initialize indexs of datasets:['./train_data/icdar2015/text_localization/test_icdar2015_label.txt'] eval model:: 95%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏ | 20/21 [00:05<00:00, 3.76it/s] [2023/03/15 09:10:02] ppocr INFO: metric eval ***** [2023/03/15 09:10:02] ppocr INFO: precision:0.0 [2023/03/15 09:10:02] ppocr INFO: recall:0.0 [2023/03/15 09:10:02] ppocr INFO: hmean:0 [2023/03/15 09:10:02] ppocr INFO: fps:3.8443568125790994 Traceback (most recent call last): File "deploy/slim/quantization/export_model.py", line 176, in main() File "deploy/slim/quantization/export_model.py", line 165, in main export_single_model(model.model_list[idx], archs[idx], File "E:\Yix\PaddleOCR-release-2.6\tools\export_model.py", line 187, in export_single_model quanter.save_quantized_model(model, save_path) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddleslim\dygraph\quant\qat.py", line 289, in save_quantized_model self.imperative_qat.save_quantized_model( File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\contrib\slim\quantization\imperative\qat.py", line 292, in save_quantized_model self._quantize_outputs.save_quantized_model(layer, path, input_spec, File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\contrib\slim\quantization\imperative\qat.py", line 500, in save_quantized_model paddle.jit.save(layer=model, path=path, input_spec=input_spec, *config) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\decorator.py", line 232, in fun return caller(func, (extras + args), kw) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\wrapped_decorator.py", line 26, in impl return wrapped_func(args, kwargs) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\dygraph\jit.py", line 649, in wrapper func(layer, path, input_spec, configs) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\decorator.py", line 232, in fun return caller(func, (extras + args), kw) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\wrapped_decorator.py", line 26, in impl return wrapped_func(*args, *kwargs) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\dygraph\base.py", line 67, in impl return func(args, kwargs) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\dygraph\jit.py", line 892, in save configs = _parse_save_configs(configs) File "D:\Program Files\anaconda3\envs\PaddleOCR-release-2.6\lib\site-packages\paddle\fluid\dygraph\jit.py", line 391, in _parse_save_configs raise ValueError( ValueError: The additional config (onnx_format) of paddle.jit.save is not supported.

LDOUBLEV commented 1 year ago

你的paddleslim版本是多少

按理说没有onnx格式参与，报错有点奇怪 ValueError: The additional config (onnx_format) of paddle.jit.save is not supported.

github-actions[bot] commented 1 year ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

PaddlePaddle / PaddleOCR

PP-OCR模型量化导出模型时报错 #9437