search

PaddlePaddle / PaddleOCR

Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)
https://paddlepaddle.github.io/PaddleOCR/
Apache License 2.0
43.9k stars 7.8k forks source link

识别训练时没有生成model日志，直接跳出 #2051

Closed 18328401960 closed 3 years ago

18328401960 commented 3 years ago

F:\PaddleOCR>python tools/train.py -c configs/rec/rec_icdar15_train.yml [2021/02/20 17:42:08] root INFO: Architecture : [2021/02/20 17:42:08] root INFO: Backbone : [2021/02/20 17:42:08] root INFO: model_name : large [2021/02/20 17:42:08] root INFO: name : MobileNetV3 [2021/02/20 17:42:08] root INFO: scale : 0.5 [2021/02/20 17:42:08] root INFO: Head : [2021/02/20 17:42:08] root INFO: fc_decay : 0 [2021/02/20 17:42:08] root INFO: name : CTCHead [2021/02/20 17:42:08] root INFO: Neck : [2021/02/20 17:42:08] root INFO: encoder_type : rnn [2021/02/20 17:42:08] root INFO: hidden_size : 96 [2021/02/20 17:42:08] root INFO: name : SequenceEncoder [2021/02/20 17:42:08] root INFO: Transform : None [2021/02/20 17:42:08] root INFO: algorithm : CRNN [2021/02/20 17:42:08] root INFO: model_type : rec [2021/02/20 17:42:08] root INFO: Eval : [2021/02/20 17:42:08] root INFO: dataset : [2021/02/20 17:42:08] root INFO: data_dir : ./train_data/ic15_data/ [2021/02/20 17:42:08] root INFO: label_file_list : ['./train_data/ic15_data/rec_gt_test.txt'] [2021/02/20 17:42:08] root INFO: name : SimpleDataSet [2021/02/20 17:42:08] root INFO: transforms : [2021/02/20 17:42:08] root INFO: DecodeImage : [2021/02/20 17:42:08] root INFO: channel_first : False [2021/02/20 17:42:08] root INFO: img_mode : BGR [2021/02/20 17:42:08] root INFO: CTCLabelEncode : None [2021/02/20 17:42:08] root INFO: RecResizeImg : [2021/02/20 17:42:08] root INFO: image_shape : [3, 32, 100] [2021/02/20 17:42:08] root INFO: KeepKeys : [2021/02/20 17:42:08] root INFO: keep_keys : ['image', 'label', 'length'] [2021/02/20 17:42:08] root INFO: loader : [2021/02/20 17:42:08] root INFO: batch_size_per_card : 256 [2021/02/20 17:42:08] root INFO: drop_last : False [2021/02/20 17:42:08] root INFO: num_workers : 4 [2021/02/20 17:42:08] root INFO: shuffle : False [2021/02/20 17:42:08] root INFO: use_shared_memory : False [2021/02/20 17:42:08] root INFO: Global : [2021/02/20 17:42:08] root INFO: cal_metric_during_train : True [2021/02/20 17:42:08] root INFO: character_dict_path : ppocr/utils/ic15_dict.txt [2021/02/20 17:42:08] root INFO: character_type : ch [2021/02/20 17:42:08] root INFO: checkpoints : None [2021/02/20 17:42:08] root INFO: debug : False [2021/02/20 17:42:08] root INFO: distributed : False [2021/02/20 17:42:08] root INFO: epoch_num : 72 [2021/02/20 17:42:08] root INFO: eval_batch_step : [0, 2000] [2021/02/20 17:42:08] root INFO: infer_img : doc/imgs_words_en/word_10.png [2021/02/20 17:42:08] root INFO: infer_mode : False [2021/02/20 17:42:08] root INFO: log_smooth_window : 20 [2021/02/20 17:42:08] root INFO: max_text_length : 25 [2021/02/20 17:42:08] root INFO: pretrained_model : ./pretrain_models/rec_mv3_none_bilstm_ctc_v2.0_train/best_accuracy [2021/02/20 17:42:08] root INFO: print_batch_step : 10 [2021/02/20 17:42:08] root INFO: save_epoch_step : 3 [2021/02/20 17:42:08] root INFO: save_inference_dir : None [2021/02/20 17:42:08] root INFO: save_model_dir : ./output/rec/ic15/ [2021/02/20 17:42:08] root INFO: use_gpu : False [2021/02/20 17:42:08] root INFO: use_space_char : False [2021/02/20 17:42:08] root INFO: use_visualdl : False [2021/02/20 17:42:08] root INFO: Loss : [2021/02/20 17:42:08] root INFO: name : CTCLoss [2021/02/20 17:42:08] root INFO: Metric : [2021/02/20 17:42:08] root INFO: main_indicator : acc [2021/02/20 17:42:08] root INFO: name : RecMetric [2021/02/20 17:42:08] root INFO: Optimizer : [2021/02/20 17:42:08] root INFO: beta1 : 0.9 [2021/02/20 17:42:08] root INFO: beta2 : 0.999 [2021/02/20 17:42:08] root INFO: lr : [2021/02/20 17:42:08] root INFO: learning_rate : 0.0005 [2021/02/20 17:42:08] root INFO: name : Adam [2021/02/20 17:42:08] root INFO: regularizer : [2021/02/20 17:42:08] root INFO: factor : 0 [2021/02/20 17:42:08] root INFO: name : L2 [2021/02/20 17:42:08] root INFO: PostProcess : [2021/02/20 17:42:08] root INFO: name : CTCLabelDecode [2021/02/20 17:42:08] root INFO: Train : [2021/02/20 17:42:08] root INFO: dataset : [2021/02/20 17:42:08] root INFO: data_dir : ./train_data/ic15_data/ [2021/02/20 17:42:08] root INFO: label_file_list : ['./train_data/ic15_data/rec_gt_train.txt'] [2021/02/20 17:42:08] root INFO: name : SimpleDataSet [2021/02/20 17:42:08] root INFO: transforms : [2021/02/20 17:42:08] root INFO: DecodeImage : [2021/02/20 17:42:08] root INFO: channel_first : False [2021/02/20 17:42:08] root INFO: img_mode : BGR [2021/02/20 17:42:08] root INFO: CTCLabelEncode : None [2021/02/20 17:42:08] root INFO: RecResizeImg : [2021/02/20 17:42:08] root INFO: image_shape : [3, 32, 100] [2021/02/20 17:42:08] root INFO: KeepKeys : [2021/02/20 17:42:08] root INFO: keep_keys : ['image', 'label', 'length'] [2021/02/20 17:42:08] root INFO: loader : [2021/02/20 17:42:08] root INFO: batch_size_per_card : 256 [2021/02/20 17:42:08] root INFO: drop_last : True [2021/02/20 17:42:08] root INFO: num_workers : 8 [2021/02/20 17:42:08] root INFO: shuffle : True [2021/02/20 17:42:08] root INFO: use_shared_memory : False [2021/02/20 17:42:08] root INFO: train with paddle 2.0.0-rc1 and device CPUPlace [2021/02/20 17:42:08] root INFO: Initialize indexs of datasets:['./train_data/ic15_data/rec_gt_train.txt'] [2021/02/20 17:42:08] root INFO: Initialize indexs of datasets:['./train_data/ic15_data/rec_gt_test.txt'] [2021/02/20 17:42:08] root INFO: load pretrained model from ['./pretrain_models/rec_mv3_none_bilstm_ctc_v2.0_train/best_accuracy'] [2021/02/20 17:42:08] root INFO: train dataloader has 17 iters, valid dataloader has 9 iters [2021/02/20 17:42:08] root INFO: During the training process, after the 0th iteration, an evaluation is run every 2000 iterations [2021/02/20 17:42:08] root INFO: Initialize indexs of datasets:['./train_data/ic15_data/rec_gt_train.txt'] [2021/02/20 17:45:05] root INFO: epoch: [1/72], iter: 10, lr: 0.000500, loss: 8.828528, acc: 0.207031, norm_edit_dis: 0.695941, reader_cost: 0.04706 s, batch_cost: 17.71910 s, samples: 2816, ips: 15.89246 image

您好，日志到这后直接跳出，没有生成model。想询问下解决办法

LDOUBLEV commented 3 years ago

[2021/02/20 17:42:08] root INFO: eval_batch_step : [0, 2000]

看你的配置是这个，表示每训练2000次才评估一次

18328401960 commented 3 years ago

[2021/02/20 17:42:08]根信息：eval_batch_step：[0，2000]

看你的配置是这个，表示每训练2000次才评估一次

[2021/02/20 17:42:08]根信息：eval_batch_step：[0，2000]

看你的配置是这个，表示每训练2000次才评估一次

好的，谢谢。