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

rec_chinese_lite_train_v2.0 训练手写体 acc 一直0 #3074

Closed wa3926 closed 2 years ago

wa3926 commented 3 years ago

paddle 版本 2.0.2 paddleOCR 是2.1 想训练一个手写体识别模型 利用的是hwdb 单字合成图片来训练的 长度已经控制 1-25 目前合成第一批 24万多一点 想试一下效果 字典 7300多字符 原始文件改动了 字典和数据集路劲 和 batch_size_per_card 没有加载预训练模型 name: Cosine 这个属性群里有人说去掉试试 也没用 训练了一夜 epoch 到9 acc 一直是0 下面贴出了 配置和 部分训练日志 求大佬指导一下思路

昨天使用预训练模型 rec_mv3_none_bilstm_ctc_v2.0_train/best_accuracy 训练了一个一夜多 到了11轮 loss 处于震荡 acc 还是0 部分训练日志在最下面

Global: use_gpu: true epoch_num: 50 log_smooth_window: 20 print_batch_step: 10 save_model_dir: ./output/rec_chinese_lite_v2.0_myhandwrite save_epoch_step: 3

evaluation is run every 5000 iterations after the 4000th iteration

eval_batch_step: [0, 2000] cal_metric_during_train: True pretrained_model: checkpoints: save_inference_dir: use_visualdl: False infer_img: doc/imgs_words/ch/word_1.jpg

for data or label process

character_dict_path: ppocr/utils/dict/myhandwrite.txt character_type: ch max_text_length: 25 infer_mode: False use_space_char: False save_res_path: ./output/rec/predicts_chinese_lite_v2.0_myhandwrite.txt

Optimizer: name: Adam beta1: 0.9 beta2: 0.999 lr: name: Cosine learning_rate: 0.001 regularizer: name: 'L2' factor: 0.00001

Architecture: model_type: rec algorithm: CRNN Transform: Backbone: name: MobileNetV3 scale: 0.5 model_name: small small_stride: [1, 2, 2, 2] Neck: name: SequenceEncoder encoder_type: rnn hidden_size: 48 Head: name: CTCHead fc_decay: 0.00001

Loss: name: CTCLoss

PostProcess: name: CTCLabelDecode

Metric: name: RecMetric main_indicator: acc

Train: dataset: name: SimpleDataSet data_dir: /paddle/data/handwrite/train_data/ label_file_list: ["/paddle/data/handwrite/train_data/train.txt"] transforms:

Eval: dataset: name: SimpleDataSet data_dir: /paddle/data/handwrite/train_data label_file_list: ["/paddle/data/handwrite/train_data/test.txt"] transforms:

下面是训练日志 [2021/06/09 08:48:15] root INFO: Architecture : [2021/06/09 08:48:15] root INFO: Backbone : [2021/06/09 08:48:15] root INFO: model_name : small [2021/06/09 08:48:15] root INFO: name : MobileNetV3 [2021/06/09 08:48:15] root INFO: scale : 0.5 [2021/06/09 08:48:15] root INFO: small_stride : [1, 2, 2, 2] [2021/06/09 08:48:15] root INFO: Head : [2021/06/09 08:48:15] root INFO: fc_decay : 1e-05 [2021/06/09 08:48:15] root INFO: name : CTCHead [2021/06/09 08:48:15] root INFO: Neck : [2021/06/09 08:48:15] root INFO: encoder_type : rnn [2021/06/09 08:48:15] root INFO: hidden_size : 48 [2021/06/09 08:48:15] root INFO: name : SequenceEncoder [2021/06/09 08:48:15] root INFO: Transform : None [2021/06/09 08:48:15] root INFO: algorithm : CRNN [2021/06/09 08:48:15] root INFO: model_type : rec [2021/06/09 08:48:15] root INFO: Eval : [2021/06/09 08:48:15] root INFO: dataset : [2021/06/09 08:48:15] root INFO: data_dir : /paddle/data/handwrite/train_data [2021/06/09 08:48:15] root INFO: label_file_list : ['/paddle/data/handwrite/train_data/test.txt'] [2021/06/09 08:48:15] root INFO: name : SimpleDataSet [2021/06/09 08:48:15] root INFO: transforms : [2021/06/09 08:48:15] root INFO: DecodeImage : [2021/06/09 08:48:15] root INFO: channel_first : False [2021/06/09 08:48:15] root INFO: img_mode : BGR [2021/06/09 08:48:15] root INFO: CTCLabelEncode : None [2021/06/09 08:48:15] root INFO: RecResizeImg : [2021/06/09 08:48:15] root INFO: image_shape : [3, 32, 320] [2021/06/09 08:48:15] root INFO: KeepKeys : [2021/06/09 08:48:15] root INFO: keep_keys : ['image', 'label', 'length'] [2021/06/09 08:48:15] root INFO: loader : [2021/06/09 08:48:15] root INFO: batch_size_per_card : 512 [2021/06/09 08:48:15] root INFO: drop_last : False [2021/06/09 08:48:15] root INFO: num_workers : 8 [2021/06/09 08:48:15] root INFO: shuffle : False [2021/06/09 08:48:15] root INFO: Global : [2021/06/09 08:48:15] root INFO: cal_metric_during_train : True [2021/06/09 08:48:15] root INFO: character_dict_path : ppocr/utils/dict/myhandwrite.txt [2021/06/09 08:48:15] root INFO: character_type : ch [2021/06/09 08:48:15] root INFO: checkpoints : None [2021/06/09 08:48:15] root INFO: debug : False [2021/06/09 08:48:15] root INFO: distributed : False [2021/06/09 08:48:15] root INFO: epoch_num : 50 [2021/06/09 08:48:15] root INFO: eval_batch_step : [0, 2000] [2021/06/09 08:48:15] root INFO: infer_img : doc/imgs_words/ch/word_1.jpg [2021/06/09 08:48:15] root INFO: infer_mode : False [2021/06/09 08:48:15] root INFO: log_smooth_window : 20 [2021/06/09 08:48:15] root INFO: max_text_length : 25 [2021/06/09 08:48:15] root INFO: pretrained_model : None [2021/06/09 08:48:15] root INFO: print_batch_step : 10 [2021/06/09 08:48:15] root INFO: save_epoch_step : 3 [2021/06/09 08:48:15] root INFO: save_inference_dir : None [2021/06/09 08:48:15] root INFO: save_model_dir : ./output/rec_chinese_lite_v2.0_myhandwrite [2021/06/09 08:48:15] root INFO: save_res_path : ./output/rec/predicts_chinese_lite_v2.0_myhandwrite.txt [2021/06/09 08:48:15] root INFO: use_gpu : True [2021/06/09 08:48:15] root INFO: use_space_char : False [2021/06/09 08:48:15] root INFO: use_visualdl : False [2021/06/09 08:48:15] root INFO: Loss : [2021/06/09 08:48:15] root INFO: name : CTCLoss [2021/06/09 08:48:15] root INFO: Metric : [2021/06/09 08:48:15] root INFO: main_indicator : acc [2021/06/09 08:48:15] root INFO: name : RecMetric [2021/06/09 08:48:15] root INFO: Optimizer : [2021/06/09 08:48:15] root INFO: beta1 : 0.9 [2021/06/09 08:48:15] root INFO: beta2 : 0.999 [2021/06/09 08:48:15] root INFO: lr : [2021/06/09 08:48:15] root INFO: learning_rate : 0.001 [2021/06/09 08:48:15] root INFO: name : Adam [2021/06/09 08:48:15] root INFO: regularizer : [2021/06/09 08:48:15] root INFO: factor : 1e-05 [2021/06/09 08:48:15] root INFO: name : L2 [2021/06/09 08:48:15] root INFO: PostProcess : [2021/06/09 08:48:15] root INFO: name : CTCLabelDecode [2021/06/09 08:48:15] root INFO: Train : [2021/06/09 08:48:15] root INFO: dataset : [2021/06/09 08:48:15] root INFO: data_dir : /paddle/data/handwrite/train_data/ [2021/06/09 08:48:15] root INFO: label_file_list : ['/paddle/data/handwrite/train_data/train.txt'] [2021/06/09 08:48:15] root INFO: name : SimpleDataSet [2021/06/09 08:48:15] root INFO: transforms : [2021/06/09 08:48:15] root INFO: DecodeImage : [2021/06/09 08:48:15] root INFO: channel_first : False [2021/06/09 08:48:15] root INFO: img_mode : BGR [2021/06/09 08:48:15] root INFO: RecAug : None [2021/06/09 08:48:15] root INFO: CTCLabelEncode : None [2021/06/09 08:48:15] root INFO: RecResizeImg : [2021/06/09 08:48:15] root INFO: image_shape : [3, 32, 320] [2021/06/09 08:48:15] root INFO: KeepKeys : [2021/06/09 08:48:15] root INFO: keep_keys : ['image', 'label', 'length'] [2021/06/09 08:48:15] root INFO: loader : [2021/06/09 08:48:15] root INFO: batch_size_per_card : 512 [2021/06/09 08:48:15] root INFO: drop_last : True [2021/06/09 08:48:15] root INFO: num_workers : 8 [2021/06/09 08:48:15] root INFO: shuffle : True [2021/06/09 08:48:15] root INFO: train with paddle 2.0.2 and device CUDAPlace(0) [2021/06/09 08:48:15] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/train.txt'] [2021/06/09 08:48:16] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/test.txt'] [2021/06/09 08:48:24] root INFO: train from scratch [2021/06/09 08:48:24] root INFO: train dataloader has 470 iters [2021/06/09 08:48:24] root INFO: valid dataloader has 118 iters [2021/06/09 08:48:24] root INFO: During the training process, after the 0th iteration, an evaluation is run every 2000 iterations [2021/06/09 08:48:24] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/train.txt'] [2021/06/09 08:52:55] root INFO: epoch: [1/50], iter: 10, lr: 0.001000, loss: 596.034973, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 17.73256 s, batch_cost: 21.79139 s, samples: 5632, ips: 25.84507 [2021/06/09 08:55:12] root INFO: epoch: [1/50], iter: 20, lr: 0.001000, loss: 477.770844, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.63379 s, batch_cost: 9.12150 s, samples: 5120, ips: 56.13113 [2021/06/09 08:57:26] root INFO: epoch: [1/50], iter: 30, lr: 0.001000, loss: 263.085541, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.98001 s, batch_cost: 9.44257 s, samples: 5120, ips: 54.22252 [2021/06/09 09:00:33] root INFO: epoch: [1/50], iter: 40, lr: 0.001000, loss: 133.886688, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.67002 s, batch_cost: 14.41033 s, samples: 5120, ips: 35.53006 [2021/06/09 09:02:53] root INFO: epoch: [1/50], iter: 50, lr: 0.001000, loss: 123.288689, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.55591 s, batch_cost: 9.03641 s, samples: 5120, ips: 56.65965 [2021/06/09 09:05:02] root INFO: epoch: [1/50], iter: 60, lr: 0.001000, loss: 121.591797, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.77928 s, batch_cost: 8.35498 s, samples: 5120, ips: 61.28078 [2021/06/09 09:07:13] root INFO: epoch: [1/50], iter: 70, lr: 0.001000, loss: 119.282883, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.00634 s, batch_cost: 8.52051 s, samples: 5120, ips: 60.09033 [2021/06/09 09:10:16] root INFO: epoch: [1/50], iter: 80, lr: 0.001000, loss: 119.507675, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.46286 s, batch_cost: 13.84152 s, samples: 5120, ips: 36.99017 [2021/06/09 09:12:25] root INFO: epoch: [1/50], iter: 90, lr: 0.001000, loss: 120.095398, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.37812 s, batch_cost: 8.72941 s, samples: 5120, ips: 58.65231 [2021/06/09 09:14:36] root INFO: epoch: [1/50], iter: 100, lr: 0.001000, loss: 118.855217, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.06551 s, batch_cost: 8.63075 s, samples: 5120, ips: 59.32276 [2021/06/09 09:16:51] root INFO: epoch: [1/50], iter: 110, lr: 0.001000, loss: 118.147430, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.86261 s, batch_cost: 8.74119 s, samples: 5120, ips: 58.57326 [2021/06/09 09:20:01] root INFO: epoch: [1/50], iter: 120, lr: 0.001000, loss: 117.627548, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.20080 s, batch_cost: 14.16490 s, samples: 5120, ips: 36.14568 [2021/06/09 09:22:04] root INFO: epoch: [1/50], iter: 130, lr: 0.001000, loss: 116.700111, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.08828 s, batch_cost: 7.77892 s, samples: 5120, ips: 65.81893 [2021/06/09 09:24:14] root INFO: epoch: [1/50], iter: 140, lr: 0.001000, loss: 116.649879, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.13201 s, batch_cost: 8.34158 s, samples: 5120, ips: 61.37924 [2021/06/09 09:26:26] root INFO: epoch: [1/50], iter: 150, lr: 0.001000, loss: 116.649879, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.79433 s, batch_cost: 8.89583 s, samples: 5120, ips: 57.55504 [2021/06/09 09:29:32] root INFO: epoch: [1/50], iter: 160, lr: 0.001000, loss: 117.215271, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.07319 s, batch_cost: 14.23928 s, samples: 5120, ips: 35.95688 [2021/06/09 09:31:50] root INFO: epoch: [1/50], iter: 170, lr: 0.001000, loss: 118.288513, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.61874 s, batch_cost: 9.21675 s, samples: 5120, ips: 55.55103 [2021/06/09 09:34:17] root INFO: epoch: [1/50], iter: 180, lr: 0.001000, loss: 118.150597, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.40194 s, batch_cost: 9.93641 s, samples: 5120, ips: 51.52768 [2021/06/09 09:36:21] root INFO: epoch: [1/50], iter: 190, lr: 0.001000, loss: 118.065857, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.80127 s, batch_cost: 8.09214 s, samples: 5120, ips: 63.27130 [2021/06/09 09:39:39] root INFO: epoch: [1/50], iter: 200, lr: 0.001000, loss: 117.881134, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 12.09171 s, batch_cost: 15.66631 s, samples: 5120, ips: 32.68159 [2021/06/09 09:42:06] root INFO: epoch: [1/50], iter: 210, lr: 0.001000, loss: 117.597321, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 7.15460 s, batch_cost: 10.55633 s, samples: 5120, ips: 48.50168 [2021/06/09 09:44:24] root INFO: epoch: [1/50], iter: 220, lr: 0.001000, loss: 117.011063, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.17669 s, batch_cost: 8.87803 s, samples: 5120, ips: 57.67045 [2021/06/09 09:46:37] root INFO: epoch: [1/50], iter: 230, lr: 0.001000, loss: 117.567291, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.15399 s, batch_cost: 8.83912 s, samples: 5120, ips: 57.92433 [2021/06/09 09:49:38] root INFO: epoch: [1/50], iter: 240, lr: 0.001000, loss: 116.428268, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 9.31889 s, batch_cost: 13.01160 s, samples: 5120, ips: 39.34951 [2021/06/09 09:51:55] root INFO: epoch: [1/50], iter: 250, lr: 0.001000, loss: 115.992920, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.68271 s, batch_cost: 9.03341 s, samples: 5120, ips: 56.67851 [2021/06/09 09:54:06] root INFO: epoch: [1/50], iter: 260, lr: 0.001000, loss: 117.065567, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.77766 s, batch_cost: 8.28685 s, samples: 5120, ips: 61.78460 [2021/06/09 09:56:20] root INFO: epoch: [1/50], iter: 270, lr: 0.001000, loss: 117.327881, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.47161 s, batch_cost: 9.05837 s, samples: 5120, ips: 56.52229 [2021/06/09 09:59:19] root INFO: epoch: [1/50], iter: 280, lr: 0.001000, loss: 117.143700, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 9.39133 s, batch_cost: 13.13456 s, samples: 5120, ips: 38.98114 [2021/06/09 10:01:35] root INFO: epoch: [1/50], iter: 290, lr: 0.001000, loss: 117.464218, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.21221 s, batch_cost: 9.02628 s, samples: 5120, ips: 56.72325 [2021/06/09 10:03:45] root INFO: epoch: [1/50], iter: 300, lr: 0.001000, loss: 116.241577, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.07484 s, batch_cost: 8.57566 s, samples: 5120, ips: 59.70386 [2021/06/09 10:05:51] root INFO: epoch: [1/50], iter: 310, lr: 0.001000, loss: 115.149841, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.59157 s, batch_cost: 7.99876 s, samples: 5120, ips: 64.00993 [2021/06/09 10:09:02] root INFO: epoch: [1/50], iter: 320, lr: 0.001000, loss: 114.557281, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.02892 s, batch_cost: 14.57455 s, samples: 5120, ips: 35.12973 [2021/06/09 10:11:19] root INFO: epoch: [1/50], iter: 330, lr: 0.001000, loss: 115.675743, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.83753 s, batch_cost: 9.44806 s, samples: 5120, ips: 54.19104 [2021/06/09 10:13:37] root INFO: epoch: [1/50], iter: 340, lr: 0.001000, loss: 116.490700, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.87762 s, batch_cost: 9.34529 s, samples: 5120, ips: 54.78694 [2021/06/09 10:15:40] root INFO: epoch: [1/50], iter: 350, lr: 0.001000, loss: 116.492851, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.37483 s, batch_cost: 7.90678 s, samples: 5120, ips: 64.75453 [2021/06/09 10:18:42] root INFO: epoch: [1/50], iter: 360, lr: 0.001000, loss: 117.851990, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.31018 s, batch_cost: 13.69805 s, samples: 5120, ips: 37.37757 [2021/06/09 10:20:51] root INFO: epoch: [1/50], iter: 370, lr: 0.001000, loss: 117.168961, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.45625 s, batch_cost: 8.70230 s, samples: 5120, ips: 58.83503 [2021/06/09 10:23:14] root INFO: epoch: [1/50], iter: 380, lr: 0.001000, loss: 115.774178, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.66665 s, batch_cost: 9.67958 s, samples: 5120, ips: 52.89486 [2021/06/09 10:25:24] root INFO: epoch: [1/50], iter: 390, lr: 0.001000, loss: 115.774178, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.91863 s, batch_cost: 8.49587 s, samples: 5120, ips: 60.26455 [2021/06/09 10:28:40] root INFO: epoch: [1/50], iter: 400, lr: 0.001000, loss: 115.508682, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.03590 s, batch_cost: 14.74334 s, samples: 5120, ips: 34.72754 [2021/06/09 10:30:56] root INFO: epoch: [1/50], iter: 410, lr: 0.001000, loss: 115.799606, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.21286 s, batch_cost: 8.75871 s, samples: 5120, ips: 58.45607 [2021/06/09 10:33:14] root INFO: epoch: [1/50], iter: 420, lr: 0.001000, loss: 115.695648, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.39109 s, batch_cost: 9.52291 s, samples: 5120, ips: 53.76507 [2021/06/09 10:35:30] root INFO: epoch: [1/50], iter: 430, lr: 0.001000, loss: 114.898773, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.19764 s, batch_cost: 9.23238 s, samples: 5120, ips: 55.45701 [2021/06/09 10:38:30] root INFO: epoch: [1/50], iter: 440, lr: 0.001000, loss: 115.507935, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 9.51051 s, batch_cost: 13.33606 s, samples: 5120, ips: 38.39216 [2021/06/09 10:40:39] root INFO: epoch: [1/50], iter: 450, lr: 0.001000, loss: 115.296814, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.03132 s, batch_cost: 8.50429 s, samples: 5120, ips: 60.20488 [2021/06/09 10:42:49] root INFO: epoch: [1/50], iter: 460, lr: 0.001000, loss: 116.092484, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.10187 s, batch_cost: 8.74840 s, samples: 5120, ips: 58.52498 [2021/06/09 10:44:52] root INFO: epoch: [1/50], iter: 469, lr: 0.001000, loss: 116.150337, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.70064 s, batch_cost: 8.58780 s, samples: 4608, ips: 53.65752 [2021/06/09 10:44:53] root INFO: save model in ./output/rec_chinese_lite_v2.0_myhandwrite/latest [2021/06/09 10:44:53] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/train.txt'] [2021/06/09 10:47:07] root INFO: epoch: [2/50], iter: 470, lr: 0.001000, loss: 116.150337, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 12.50566 s, batch_cost: 12.91081 s, samples: 512, ips: 3.96567 [2021/06/09 10:49:16] root INFO: epoch: [2/50], iter: 480, lr: 0.001000, loss: 115.530632, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.82029 s, batch_cost: 8.30905 s, samples: 5120, ips: 61.61958 [2021/06/09 10:51:31] root INFO: epoch: [2/50], iter: 490, lr: 0.001000, loss: 115.530632, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.95351 s, batch_cost: 9.31322 s, samples: 5120, ips: 54.97561 [2021/06/09 10:54:00] root INFO: epoch: [2/50], iter: 500, lr: 0.001000, loss: 117.003006, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.80905 s, batch_cost: 10.17362 s, samples: 5120, ips: 50.32626 [2021/06/09 10:57:08] root INFO: epoch: [2/50], iter: 510, lr: 0.001000, loss: 116.701233, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.30436 s, batch_cost: 14.67373 s, samples: 5120, ips: 34.89229 [2021/06/09 10:59:20] root INFO: epoch: [2/50], iter: 520, lr: 0.001000, loss: 116.565788, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.02843 s, batch_cost: 8.52582 s, samples: 5120, ips: 60.05289 [2021/06/09 11:01:27] root INFO: epoch: [2/50], iter: 530, lr: 0.001000, loss: 116.565788, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 3.94122 s, batch_cost: 7.90919 s, samples: 5120, ips: 64.73484 [2021/06/09 11:03:41] root INFO: epoch: [2/50], iter: 540, lr: 0.001000, loss: 115.881828, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.92106 s, batch_cost: 8.56406 s, samples: 5120, ips: 59.78476 [2021/06/09 11:06:44] root INFO: epoch: [2/50], iter: 550, lr: 0.001000, loss: 117.548264, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.19929 s, batch_cost: 13.74965 s, samples: 5120, ips: 37.23732 [2021/06/09 11:09:01] root INFO: epoch: [2/50], iter: 560, lr: 0.001000, loss: 117.881088, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.74696 s, batch_cost: 9.41013 s, samples: 5120, ips: 54.40943 [2021/06/09 11:11:17] root INFO: epoch: [2/50], iter: 570, lr: 0.001000, loss: 117.324455, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.81892 s, batch_cost: 8.31224 s, samples: 5120, ips: 61.59589 [2021/06/09 11:13:22] root INFO: epoch: [2/50], iter: 580, lr: 0.001000, loss: 116.340477, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.45636 s, batch_cost: 8.08446 s, samples: 5120, ips: 63.33139 [2021/06/09 11:16:35] root INFO: epoch: [2/50], iter: 590, lr: 0.001000, loss: 117.224304, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.13230 s, batch_cost: 14.72520 s, samples: 5120, ips: 34.77032 [2021/06/09 11:18:51] root INFO: epoch: [2/50], iter: 600, lr: 0.001000, loss: 117.316528, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.65136 s, batch_cost: 8.97453 s, samples: 5120, ips: 57.05033 [2021/06/09 11:21:17] root INFO: epoch: [2/50], iter: 610, lr: 0.001000, loss: 117.552521, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.98787 s, batch_cost: 10.28851 s, samples: 5120, ips: 49.76424 [2021/06/09 11:23:27] root INFO: epoch: [2/50], iter: 620, lr: 0.001000, loss: 117.368279, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.88149 s, batch_cost: 8.60193 s, samples: 5120, ips: 59.52153 [2021/06/09 11:26:20] root INFO: epoch: [2/50], iter: 630, lr: 0.001000, loss: 116.263893, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 8.81072 s, batch_cost: 12.61057 s, samples: 5120, ips: 40.60087 [2021/06/09 11:28:37] root INFO: epoch: [2/50], iter: 640, lr: 0.001000, loss: 116.411758, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.82287 s, batch_cost: 9.63205 s, samples: 5120, ips: 53.15589 [2021/06/09 11:30:41] root INFO: epoch: [2/50], iter: 650, lr: 0.001000, loss: 116.529358, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.14229 s, batch_cost: 7.80108 s, samples: 5120, ips: 65.63196 [2021/06/09 11:33:00] root INFO: epoch: [2/50], iter: 660, lr: 0.001000, loss: 116.579651, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.85043 s, batch_cost: 9.45269 s, samples: 5120, ips: 54.16446 [2021/06/09 11:35:55] root INFO: epoch: [2/50], iter: 670, lr: 0.001000, loss: 117.045799, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 9.57002 s, batch_cost: 13.20515 s, samples: 5120, ips: 38.77276 [2021/06/09 11:38:20] root INFO: epoch: [2/50], iter: 680, lr: 0.001000, loss: 116.000267, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 7.06219 s, batch_cost: 10.47598 s, samples: 5120, ips: 48.87371 [2021/06/09 11:40:32] root INFO: epoch: [2/50], iter: 690, lr: 0.001000, loss: 115.417763, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.71175 s, batch_cost: 9.01625 s, samples: 5120, ips: 56.78635 [2021/06/09 11:42:36] root INFO: epoch: [2/50], iter: 700, lr: 0.001000, loss: 115.720116, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.84458 s, batch_cost: 8.32725 s, samples: 5120, ips: 61.48489 [2021/06/09 11:45:35] root INFO: epoch: [2/50], iter: 710, lr: 0.001000, loss: 116.141479, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.01915 s, batch_cost: 13.46515 s, samples: 5120, ips: 38.02408 [2021/06/09 11:48:00] root INFO: epoch: [2/50], iter: 720, lr: 0.001000, loss: 114.995476, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.04107 s, batch_cost: 9.88394 s, samples: 5120, ips: 51.80122 [2021/06/09 11:50:15] root INFO: epoch: [2/50], iter: 730, lr: 0.001000, loss: 115.429016, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.44131 s, batch_cost: 8.86453 s, samples: 5120, ips: 57.75830 [2021/06/09 11:52:39] root INFO: epoch: [2/50], iter: 740, lr: 0.001000, loss: 116.945557, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.78044 s, batch_cost: 10.17663 s, samples: 5120, ips: 50.31134 [2021/06/09 11:55:18] root INFO: epoch: [2/50], iter: 750, lr: 0.001000, loss: 116.471558, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 8.11596 s, batch_cost: 11.52092 s, samples: 5120, ips: 44.44091 [2021/06/09 11:58:05] root INFO: epoch: [2/50], iter: 760, lr: 0.001000, loss: 115.156113, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 8.78884 s, batch_cost: 12.13843 s, samples: 5120, ips: 42.18009 [2021/06/09 12:00:18] root INFO: epoch: [2/50], iter: 770, lr: 0.001000, loss: 116.445877, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.04173 s, batch_cost: 8.67772 s, samples: 5120, ips: 59.00166 [2021/06/09 12:02:19] root INFO: epoch: [2/50], iter: 780, lr: 0.001000, loss: 115.967712, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.02929 s, batch_cost: 7.47180 s, samples: 5120, ips: 68.52428 [2021/06/09 12:04:51] root INFO: epoch: [2/50], iter: 790, lr: 0.001000, loss: 115.939537, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 7.11832 s, batch_cost: 10.43809 s, samples: 5120, ips: 49.05110 [2021/06/09 12:07:41] root INFO: epoch: [2/50], iter: 800, lr: 0.001000, loss: 116.374748, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 8.96263 s, batch_cost: 12.53962 s, samples: 5120, ips: 40.83060 [2021/06/09 12:09:43] root INFO: epoch: [2/50], iter: 810, lr: 0.001000, loss: 114.210091, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 3.56057 s, batch_cost: 7.39056 s, samples: 5120, ips: 69.27761 [2021/06/09 12:12:03] root INFO: epoch: [2/50], iter: 820, lr: 0.001000, loss: 115.811317, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.08630 s, batch_cost: 9.40550 s, samples: 5120, ips: 54.43626 [2021/06/09 12:14:33] root INFO: epoch: [2/50], iter: 830, lr: 0.001000, loss: 116.169189, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 7.08321 s, batch_cost: 10.96510 s, samples: 5120, ips: 46.69360 [2021/06/09 12:17:30] root INFO: epoch: [2/50], iter: 840, lr: 0.001000, loss: 116.410385, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 9.88089 s, batch_cost: 13.18711 s, samples: 5120, ips: 38.82578 [2021/06/09 12:19:42] root INFO: epoch: [2/50], iter: 850, lr: 0.001000, loss: 116.306000, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.73768 s, batch_cost: 8.92241 s, samples: 5120, ips: 57.38360 [2021/06/09 12:21:42] root INFO: epoch: [2/50], iter: 860, lr: 0.001000, loss: 115.904099, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 3.86260 s, batch_cost: 7.57478 s, samples: 5120, ips: 67.59269 [2021/06/09 12:24:05] root INFO: epoch: [2/50], iter: 870, lr: 0.001000, loss: 115.682106, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.88506 s, batch_cost: 9.55290 s, samples: 5120, ips: 53.59631 [2021/06/09 12:26:47] root INFO: epoch: [2/50], iter: 880, lr: 0.001000, loss: 115.876770, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 8.62108 s, batch_cost: 11.86781 s, samples: 5120, ips: 43.14190 [2021/06/09 12:29:08] root INFO: epoch: [2/50], iter: 890, lr: 0.001000, loss: 116.181664, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.43322 s, batch_cost: 9.67069 s, samples: 5120, ips: 52.94350 [2021/06/09 12:31:18] root INFO: epoch: [2/50], iter: 900, lr: 0.001000, loss: 116.933632, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.36430 s, batch_cost: 8.61088 s, samples: 5120, ips: 59.45965 [2021/06/09 12:33:50] root INFO: epoch: [2/50], iter: 910, lr: 0.001000, loss: 116.092583, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 7.82389 s, batch_cost: 11.01281 s, samples: 5120, ips: 46.49132 [2021/06/09 12:36:27] root INFO: epoch: [2/50], iter: 920, lr: 0.001000, loss: 116.092583, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 8.01518 s, batch_cost: 11.30280 s, samples: 5120, ips: 45.29850 [2021/06/09 12:38:54] root INFO: epoch: [2/50], iter: 930, lr: 0.001000, loss: 116.121231, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.94875 s, batch_cost: 10.51294 s, samples: 5120, ips: 48.70187 [2021/06/09 12:40:49] root INFO: epoch: [2/50], iter: 939, lr: 0.001000, loss: 115.638832, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.31165 s, batch_cost: 7.97128 s, samples: 4608, ips: 57.80753 [2021/06/09 12:40:50] root INFO: save model in ./output/rec_chinese_lite_v2.0_myhandwrite/latest [2021/06/09 12:40:50] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/train.txt'] [2021/06/09 12:42:58] root INFO: epoch: [3/50], iter: 940, lr: 0.001000, loss: 115.288422, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 12.01046 s, batch_cost: 12.30474 s, samples: 512, ips: 4.16100 [2021/06/09 12:45:15] root INFO: epoch: [3/50], iter: 950, lr: 0.001000, loss: 116.310486, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.91265 s, batch_cost: 9.34478 s, samples: 5120, ips: 54.78994 [2021/06/09 12:47:31] root INFO: epoch: [3/50], iter: 960, lr: 0.001000, loss: 116.493134, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.56483 s, batch_cost: 9.10960 s, samples: 5120, ips: 56.20443 [2021/06/09 12:49:44] root INFO: epoch: [3/50], iter: 970, lr: 0.001000, loss: 116.493134, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.86894 s, batch_cost: 9.11881 s, samples: 5120, ips: 56.14768 [2021/06/09 12:52:58] root INFO: epoch: [3/50], iter: 980, lr: 0.001000, loss: 116.718796, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.27137 s, batch_cost: 14.88105 s, samples: 5120, ips: 34.40617 [2021/06/09 12:55:30] root INFO: epoch: [3/50], iter: 990, lr: 0.001000, loss: 116.812416, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 7.01358 s, batch_cost: 10.50748 s, samples: 5120, ips: 48.72719 [2021/06/09 12:57:30] root INFO: epoch: [3/50], iter: 1000, lr: 0.001000, loss: 115.566010, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.49768 s, batch_cost: 7.76864 s, samples: 5120, ips: 65.90599 [2021/06/09 12:59:50] root INFO: epoch: [3/50], iter: 1010, lr: 0.001000, loss: 114.928543, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.22187 s, batch_cost: 9.52842 s, samples: 5120, ips: 53.73401 [2021/06/09 13:03:05] root INFO: epoch: [3/50], iter: 1020, lr: 0.001000, loss: 114.777298, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.42577 s, batch_cost: 14.99884 s, samples: 5120, ips: 34.13598 [2021/06/09 13:05:22] root INFO: epoch: [3/50], iter: 1030, lr: 0.001000, loss: 114.795303, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.42139 s, batch_cost: 9.06739 s, samples: 5120, ips: 56.46607 [2021/06/09 13:07:18] root INFO: epoch: [3/50], iter: 1040, lr: 0.001000, loss: 114.859802, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.41018 s, batch_cost: 7.51735 s, samples: 5120, ips: 68.10913 [2021/06/09 13:09:31] root INFO: epoch: [3/50], iter: 1050, lr: 0.001000, loss: 114.987869, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.20294 s, batch_cost: 8.80066 s, samples: 5120, ips: 58.17748 [2021/06/09 13:12:37] root INFO: epoch: [3/50], iter: 1060, lr: 0.001000, loss: 115.419724, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.66784 s, batch_cost: 14.21446 s, samples: 5120, ips: 36.01967 [2021/06/09 13:15:08] root INFO: epoch: [3/50], iter: 1070, lr: 0.001000, loss: 115.792023, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 7.39688 s, batch_cost: 10.80994 s, samples: 5120, ips: 47.36380 [2021/06/09 13:17:22] root INFO: epoch: [3/50], iter: 1080, lr: 0.001000, loss: 116.131577, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.00762 s, batch_cost: 8.82650 s, samples: 5120, ips: 58.00714 [2021/06/09 13:19:39] root INFO: epoch: [3/50], iter: 1090, lr: 0.001000, loss: 114.802872, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.87055 s, batch_cost: 8.67314 s, samples: 5120, ips: 59.03280 [2021/06/09 13:22:51] root INFO: epoch: [3/50], iter: 1100, lr: 0.001000, loss: 115.783508, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.46716 s, batch_cost: 14.77010 s, samples: 5120, ips: 34.66462 [2021/06/09 13:25:12] root INFO: epoch: [3/50], iter: 1110, lr: 0.001000, loss: 116.581451, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.89029 s, batch_cost: 9.40080 s, samples: 5120, ips: 54.46346 [2021/06/09 13:27:32] root INFO: epoch: [3/50], iter: 1120, lr: 0.001000, loss: 116.497467, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.23917 s, batch_cost: 9.77569 s, samples: 5120, ips: 52.37483 [2021/06/09 13:29:41] root INFO: epoch: [3/50], iter: 1130, lr: 0.001000, loss: 115.977570, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.50356 s, batch_cost: 8.59447 s, samples: 5120, ips: 59.57319 [2021/06/09 13:33:02] root INFO: epoch: [3/50], iter: 1140, lr: 0.001000, loss: 115.682602, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 12.43427 s, batch_cost: 15.60687 s, samples: 5120, ips: 32.80607 [2021/06/09 13:35:09] root INFO: epoch: [3/50], iter: 1150, lr: 0.001000, loss: 115.415108, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.11005 s, batch_cost: 8.53526 s, samples: 5120, ips: 59.98649 [2021/06/09 13:37:19] root INFO: epoch: [3/50], iter: 1160, lr: 0.001000, loss: 115.001724, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.32758 s, batch_cost: 8.85395 s, samples: 5120, ips: 57.82729 [2021/06/09 13:39:33] root INFO: epoch: [3/50], iter: 1170, lr: 0.001000, loss: 115.251877, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.65283 s, batch_cost: 8.89875 s, samples: 5120, ips: 57.53619 [2021/06/09 13:42:48] root INFO: epoch: [3/50], iter: 1180, lr: 0.001000, loss: 115.301270, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.70592 s, batch_cost: 14.93867 s, samples: 5120, ips: 34.27346 [2021/06/09 13:44:55] root INFO: epoch: [3/50], iter: 1190, lr: 0.001000, loss: 115.372818, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.31443 s, batch_cost: 7.83429 s, samples: 5120, ips: 65.35368 [2021/06/09 13:47:11] root INFO: epoch: [3/50], iter: 1200, lr: 0.001000, loss: 115.850899, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.19803 s, batch_cost: 8.83596 s, samples: 5120, ips: 57.94501 [2021/06/09 13:49:26] root INFO: epoch: [3/50], iter: 1210, lr: 0.001000, loss: 114.891418, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.86167 s, batch_cost: 9.05075 s, samples: 5120, ips: 56.56991 [2021/06/09 13:52:43] root INFO: epoch: [3/50], iter: 1220, lr: 0.001000, loss: 113.906822, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 12.14878 s, batch_cost: 15.29807 s, samples: 5120, ips: 33.46827 [2021/06/09 13:54:59] root INFO: epoch: [3/50], iter: 1230, lr: 0.001000, loss: 115.019508, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.49713 s, batch_cost: 8.92254 s, samples: 5120, ips: 57.38276 [2021/06/09 13:57:11] root INFO: epoch: [3/50], iter: 1240, lr: 0.001000, loss: 115.296913, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.46911 s, batch_cost: 8.47450 s, samples: 5120, ips: 60.41656 [2021/06/09 13:59:15] root INFO: epoch: [3/50], iter: 1250, lr: 0.001000, loss: 115.197701, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 4.48975 s, batch_cost: 7.96059 s, samples: 5120, ips: 64.31681 [2021/06/09 14:02:23] root INFO: epoch: [3/50], iter: 1260, lr: 0.001000, loss: 115.826477, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.24289 s, batch_cost: 14.14691 s, samples: 5120, ips: 36.19166 [2021/06/09 14:04:36] root INFO: epoch: [3/50], iter: 1270, lr: 0.001000, loss: 116.233437, acc: 0.000000, norm_edit_dis: 0.000078, reader_cost: 5.78990 s, batch_cost: 8.79996 s, samples: 5120, ips: 58.18206 [2021/06/09 14:06:51] root INFO: epoch: [3/50], iter: 1280, lr: 0.001000, loss: 113.827774, acc: 0.000000, norm_edit_dis: 0.000078, reader_cost: 5.59293 s, batch_cost: 9.24499 s, samples: 5120, ips: 55.38135 [2021/06/09 14:09:12] root INFO: epoch: [3/50], iter: 1290, lr: 0.001000, loss: 113.856422, acc: 0.000000, norm_edit_dis: 0.000039, reader_cost: 5.71760 s, batch_cost: 9.47898 s, samples: 5120, ips: 54.01422 [2021/06/09 14:12:08] root INFO: epoch: [3/50], iter: 1300, lr: 0.001000, loss: 115.109230, acc: 0.000000, norm_edit_dis: 0.000039, reader_cost: 9.93765 s, batch_cost: 13.44209 s, samples: 5120, ips: 38.08930 [2021/06/09 14:14:28] root INFO: epoch: [3/50], iter: 1310, lr: 0.001000, loss: 114.801552, acc: 0.000000, norm_edit_dis: 0.000089, reader_cost: 5.48779 s, batch_cost: 8.86167 s, samples: 5120, ips: 57.77689 [2021/06/09 14:16:46] root INFO: epoch: [3/50], iter: 1320, lr: 0.001000, loss: 115.284531, acc: 0.000000, norm_edit_dis: 0.000104, reader_cost: 5.92058 s, batch_cost: 9.39520 s, samples: 5120, ips: 54.49592 [2021/06/09 14:18:42] root INFO: epoch: [3/50], iter: 1330, lr: 0.001000, loss: 114.990601, acc: 0.000000, norm_edit_dis: 0.000081, reader_cost: 4.27600 s, batch_cost: 7.36451 s, samples: 5120, ips: 69.52258 [2021/06/09 14:21:56] root INFO: epoch: [3/50], iter: 1340, lr: 0.001000, loss: 114.313988, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.23778 s, batch_cost: 14.59404 s, samples: 5120, ips: 35.08282 [2021/06/09 14:24:09] root INFO: epoch: [3/50], iter: 1350, lr: 0.001000, loss: 115.315887, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.34735 s, batch_cost: 8.81538 s, samples: 5120, ips: 58.08028 [2021/06/09 14:26:29] root INFO: epoch: [3/50], iter: 1360, lr: 0.001000, loss: 115.961136, acc: 0.000000, norm_edit_dis: 0.000039, reader_cost: 6.21968 s, batch_cost: 9.76476 s, samples: 5120, ips: 52.43344 [2021/06/09 14:28:44] root INFO: epoch: [3/50], iter: 1370, lr: 0.001000, loss: 114.718689, acc: 0.000000, norm_edit_dis: 0.000086, reader_cost: 4.81764 s, batch_cost: 8.83388 s, samples: 5120, ips: 57.95867 [2021/06/09 14:31:44] root INFO: epoch: [3/50], iter: 1380, lr: 0.001000, loss: 113.873207, acc: 0.000000, norm_edit_dis: 0.000039, reader_cost: 10.12557 s, batch_cost: 13.54966 s, samples: 5120, ips: 37.78692 [2021/06/09 14:34:05] root INFO: epoch: [3/50], iter: 1390, lr: 0.001000, loss: 114.733200, acc: 0.000000, norm_edit_dis: 0.000091, reader_cost: 5.69385 s, batch_cost: 9.15760 s, samples: 5120, ips: 55.90984 [2021/06/09 14:36:08] root INFO: epoch: [3/50], iter: 1400, lr: 0.001000, loss: 114.442352, acc: 0.000000, norm_edit_dis: 0.000110, reader_cost: 5.28842 s, batch_cost: 8.35195 s, samples: 5120, ips: 61.30308 [2021/06/09 14:38:08] root INFO: epoch: [3/50], iter: 1409, lr: 0.001000, loss: 115.075340, acc: 0.000000, norm_edit_dis: 0.000109, reader_cost: 5.60382 s, batch_cost: 8.60172 s, samples: 4608, ips: 53.57070

昨天使用预训练模型 rec_mv3_none_bilstm_ctc_v2.0_train/best_accuracy 训练了一个一夜多 到了11轮 loss 处于震荡 acc 还是0 下面是部分训练日志 [2021/06/10 03:09:17] root INFO: train with paddle 2.0.2 and device CUDAPlace(0) [2021/06/10 03:09:17] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/train.txt'] [2021/06/10 03:09:18] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/test.txt'] [2021/06/10 03:09:24] root INFO: load pretrained model from ['/paddle/PaddleOCR/pretrain_models/rec_mv3_none_bilstm_ctc_v2.0_train/best_accuracy'] [2021/06/10 03:09:24] root INFO: train dataloader has 470 iters [2021/06/10 03:09:24] root INFO: valid dataloader has 118 iters [2021/06/10 03:09:24] root INFO: During the training process, after the 0th iteration, an evaluation is run every 2000 iterations [2021/06/10 03:09:25] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/train.txt'] [2021/06/10 03:13:54] root INFO: epoch: [1/50], iter: 10, lr: 0.001000, loss: 595.696533, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 18.93526 s, batch_cost: 22.31017 s, samples: 5632, ips: 25.24410 [2021/06/10 03:16:07] root INFO: epoch: [1/50], iter: 20, lr: 0.001000, loss: 478.828186, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.34329 s, batch_cost: 8.75869 s, samples: 5120, ips: 58.45622 [2021/06/10 03:18:22] root INFO: epoch: [1/50], iter: 30, lr: 0.001000, loss: 266.721802, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 6.29737 s, batch_cost: 9.39510 s, samples: 5120, ips: 54.49649 [2021/06/10 03:21:24] root INFO: epoch: [1/50], iter: 40, lr: 0.001000, loss: 135.067291, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 10.62354 s, batch_cost: 13.89763 s, samples: 5120, ips: 36.84082 [2021/06/10 03:23:35] root INFO: epoch: [1/50], iter: 50, lr: 0.001000, loss: 123.361359, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.93160 s, batch_cost: 9.02072 s, samples: 5120, ips: 56.75824 [2021/06/10 03:25:46] root INFO: epoch: [1/50], iter: 60, lr: 0.001000, loss: 121.627274, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.78582 s, batch_cost: 8.83760 s, samples: 5120, ips: 57.93429 [2021/06/10 03:27:51] root INFO: epoch: [1/50], iter: 70, lr: 0.001000, loss: 119.333450, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.11343 s, batch_cost: 8.30963 s, samples: 5120, ips: 61.61528 [2021/06/10 03:30:58] root INFO: epoch: [1/50], iter: 80, lr: 0.001000, loss: 119.528259, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 11.59923 s, batch_cost: 14.64403 s, samples: 5120, ips: 34.96305 [2021/06/10 03:33:10] root INFO: epoch: [1/50], iter: 90, lr: 0.001000, loss: 119.921906, acc: 0.000000, norm_edit_dis: 0.000000, reader_cost: 5.44023 s, batch_cost: 8.78971 s, samples [2021/06/10 23:10:41] root INFO: epoch: [11/50], iter: 4850, lr: 0.000899, loss: 109.828331, acc: 0.000000, norm_edit_dis: 0.000945, reader_cost: 5.13419 s, batch_cost: 8.37138 s, samples: 5120, ips: 61.16079 [2021/06/10 23:12:50] root INFO: epoch: [11/50], iter: 4860, lr: 0.000899, loss: 110.466591, acc: 0.000000, norm_edit_dis: 0.000923, reader_cost: 5.87794 s, batch_cost: 8.82646 s, samples: 5120, ips: 58.00738 [2021/06/10 23:15:15] root INFO: epoch: [11/50], iter: 4870, lr: 0.000898, loss: 109.903549, acc: 0.000000, norm_edit_dis: 0.000832, reader_cost: 7.18269 s, batch_cost: 10.34768 s, samples: 5120, ips: 49.47968 [2021/06/10 23:18:12] root INFO: epoch: [11/50], iter: 4880, lr: 0.000898, loss: 110.078484, acc: 0.000000, norm_edit_dis: 0.000706, reader_cost: 10.15888 s, batch_cost: 13.50406 s, samples: 5120, ips: 37.91454 [2021/06/10 23:20:09] root INFO: epoch: [11/50], iter: 4890, lr: 0.000897, loss: 110.279884, acc: 0.000000, norm_edit_dis: 0.000777, reader_cost: 4.14654 s, batch_cost: 7.37372 s, samples: 5120, ips: 69.43576 [2021/06/10 23:22:28] root INFO: epoch: [11/50], iter: 4900, lr: 0.000897, loss: 110.279884, acc: 0.000000, norm_edit_dis: 0.000742, reader_cost: 6.83937 s, batch_cost: 9.77803 s, samples: 5120, ips: 52.36226 [2021/06/10 23:24:56] root INFO: epoch: [11/50], iter: 4910, lr: 0.000896, loss: 111.201164, acc: 0.000000, norm_edit_dis: 0.000757, reader_cost: 7.39534 s, batch_cost: 10.44557 s, samples: 5120, ips: 49.01600 [2021/06/10 23:27:48] root INFO: epoch: [11/50], iter: 4920, lr: 0.000896, loss: 111.236893, acc: 0.000000, norm_edit_dis: 0.000819, reader_cost: 9.64196 s, batch_cost: 13.12299 s, samples: 5120, ips: 39.01549 [2021/06/10 23:29:57] root INFO: epoch: [11/50], iter: 4930, lr: 0.000896, loss: 109.496216, acc: 0.000000, norm_edit_dis: 0.001028, reader_cost: 5.78394 s, batch_cost: 8.82122 s, samples: 5120, ips: 58.04186 [2021/06/10 23:32:08] root INFO: epoch: [11/50], iter: 4940, lr: 0.000895, loss: 109.282547, acc: 0.000000, norm_edit_dis: 0.001011, reader_cost: 5.76773 s, batch_cost: 9.05767 s, samples: 5120, ips: 56.52668 [2021/06/10 23:34:25] root INFO: epoch: [11/50], iter: 4950, lr: 0.000895, loss: 109.984756, acc: 0.000000, norm_edit_dis: 0.000781, reader_cost: 6.52903 s, batch_cost: 9.71990 s, samples: 5120, ips: 52.67546 [2021/06/10 23:37:18] root INFO: epoch: [11/50], iter: 4960, lr: 0.000894, loss: 109.569656, acc: 0.000000, norm_edit_dis: 0.000781, reader_cost: 9.79899 s, batch_cost: 12.85416 s, samples: 5120, ips: 39.83146 [2021/06/10 23:39:31] root INFO: epoch: [11/50], iter: 4970, lr: 0.000894, loss: 110.466972, acc: 0.000000, norm_edit_dis: 0.000764, reader_cost: 5.93452 s, batch_cost: 8.92482 s, samples: 5120, ips: 57.36809 [2021/06/10 23:41:45] root INFO: epoch: [11/50], iter: 4980, lr: 0.000894, loss: 110.256889, acc: 0.000000, norm_edit_dis: 0.000851, reader_cost: 6.16786 s, batch_cost: 9.26160 s, samples: 5120, ips: 55.28204 [2021/06/10 23:43:57] root INFO: epoch: [11/50], iter: 4990, lr: 0.000893, loss: 109.572327, acc: 0.000000, norm_edit_dis: 0.000656, reader_cost: 5.56337 s, batch_cost: 9.00928 s, samples: 5120, ips: 56.83028 [2021/06/10 23:47:01] root INFO: epoch: [11/50], iter: 5000, lr: 0.000893, loss: 109.184250, acc: 0.000000, norm_edit_dis: 0.000781, reader_cost: 10.75321 s, batch_cost: 14.10866 s, samples: 5120, ips: 36.28978 [2021/06/10 23:49:14] root INFO: epoch: [11/50], iter: 5010, lr: 0.000892, loss: 109.344101, acc: 0.000000, norm_edit_dis: 0.000966, reader_cost: 5.71265 s, batch_cost: 8.96031 s, samples: 5120, ips: 57.14089 [2021/06/10 23:51:18] root INFO: epoch: [11/50], iter: 5020, lr: 0.000892, loss: 109.854294, acc: 0.000000, norm_edit_dis: 0.000967, reader_cost: 4.89632 s, batch_cost: 8.28509 s, samples: 5120, ips: 61.79778 [2021/06/10 23:53:32] root INFO: epoch: [11/50], iter: 5030, lr: 0.000892, loss: 109.324654, acc: 0.000000, norm_edit_dis: 0.000686, reader_cost: 5.80450 s, batch_cost: 8.97446 s, samples: 5120, ips: 57.05077 [2021/06/10 23:56:25] root INFO: epoch: [11/50], iter: 5040, lr: 0.000891, loss: 109.579964, acc: 0.000000, norm_edit_dis: 0.000864, reader_cost: 10.17027 s, batch_cost: 13.27097 s, samples: 5120, ips: 38.58045 [2021/06/10 23:58:38] root INFO: epoch: [11/50], iter: 5050, lr: 0.000891, loss: 109.345673, acc: 0.000000, norm_edit_dis: 0.001117, reader_cost: 6.31339 s, batch_cost: 9.38905 s, samples: 5120, ips: 54.53162 [2021/06/11 00:00:52] root INFO: epoch: [11/50], iter: 5060, lr: 0.000890, loss: 109.570847, acc: 0.000000, norm_edit_dis: 0.000811, reader_cost: 5.92575 s, batch_cost: 9.00443 s, samples: 5120, ips: 56.86087 [2021/06/11 00:03:06] root INFO: epoch: [11/50], iter: 5070, lr: 0.000890, loss: 110.711823, acc: 0.000000, norm_edit_dis: 0.000850, reader_cost: 5.92734 s, batch_cost: 9.12112 s, samples: 5120, ips: 56.13346 [2021/06/11 00:06:21] root INFO: epoch: [11/50], iter: 5080, lr: 0.000889, loss: 110.216118, acc: 0.000000, norm_edit_dis: 0.000765, reader_cost: 12.08127 s, batch_cost: 15.42015 s, samples: 5120, ips: 33.20331 [2021/06/11 00:08:28] root INFO: epoch: [11/50], iter: 5090, lr: 0.000889, loss: 109.097977, acc: 0.000000, norm_edit_dis: 0.000835, reader_cost: 5.49313 s, batch_cost: 8.49210 s, samples: 5120, ips: 60.29132 [2021/06/11 00:10:35] root INFO: epoch: [11/50], iter: 5100, lr: 0.000889, loss: 109.513382, acc: 0.000000, norm_edit_dis: 0.000911, reader_cost: 4.82826 s, batch_cost: 8.42433 s, samples: 5120, ips: 60.77638 [2021/06/11 00:12:44] root INFO: epoch: [11/50], iter: 5110, lr: 0.000888, loss: 109.808273, acc: 0.000000, norm_edit_dis: 0.000820, reader_cost: 5.93481 s, batch_cost: 8.89376 s, samples: 5120, ips: 57.56847 [2021/06/11 00:16:06] root INFO: epoch: [11/50], iter: 5120, lr: 0.000888, loss: 109.807137, acc: 0.000000, norm_edit_dis: 0.000755, reader_cost: 12.85068 s, batch_cost: 16.02436 s, samples: 5120, ips: 31.95135 [2021/06/11 00:18:16] root INFO: epoch: [11/50], iter: 5130, lr: 0.000887, loss: 110.194809, acc: 0.000000, norm_edit_dis: 0.000761, reader_cost: 5.63601 s, batch_cost: 8.68520 s, samples: 5120, ips: 58.95083 [2021/06/11 00:20:25] root INFO: epoch: [11/50], iter: 5140, lr: 0.000887, loss: 109.842293, acc: 0.000000, norm_edit_dis: 0.000748, reader_cost: 5.47446 s, batch_cost: 8.81515 s, samples: 5120, ips: 58.08183 [2021/06/11 00:22:36] root INFO: epoch: [11/50], iter: 5150, lr: 0.000887, loss: 109.842293, acc: 0.000000, norm_edit_dis: 0.000820, reader_cost: 5.84214 s, batch_cost: 8.92297 s, samples: 5120, ips: 57.38000

littletomatodonkey commented 3 years ago

epoch改为500试下吧，我这边训6624个字符的训练集，500epoch的情况下，在20epoch的时候会有大概20%的acc

wa3926 commented 3 years ago

@littletomatodonkey 经常打扰大佬 非常抱歉 我把字典换成 ppocr_keys_v1 重新弄数据 然后再尝试下

wa3926 commented 3 years ago

@littletomatodonkey 大佬现在我准备了500W左右的数据 在之前陆陆续续训练的模型基础之上训练 显卡v100 32G显存 我估计了一下一轮大概30多小时 这训练是要按照年计算吗.........请问大佬 单卡情况下有加快训练的办法吗 [2021/06/30 10:05:44] root INFO: Architecture : [2021/06/30 10:05:44] root INFO: Backbone : [2021/06/30 10:05:44] root INFO: model_name : small [2021/06/30 10:05:44] root INFO: name : MobileNetV3 [2021/06/30 10:05:44] root INFO: scale : 0.5 [2021/06/30 10:05:44] root INFO: small_stride : [1, 2, 2, 2] [2021/06/30 10:05:44] root INFO: Head : [2021/06/30 10:05:44] root INFO: fc_decay : 1e-05 [2021/06/30 10:05:44] root INFO: name : CTCHead [2021/06/30 10:05:44] root INFO: Neck : [2021/06/30 10:05:44] root INFO: encoder_type : rnn [2021/06/30 10:05:44] root INFO: hidden_size : 48 [2021/06/30 10:05:44] root INFO: name : SequenceEncoder [2021/06/30 10:05:44] root INFO: Transform : None [2021/06/30 10:05:44] root INFO: algorithm : CRNN [2021/06/30 10:05:44] root INFO: model_type : rec [2021/06/30 10:05:44] root INFO: Eval : [2021/06/30 10:05:44] root INFO: dataset : [2021/06/30 10:05:44] root INFO: data_dir : /paddle/data/handwrite/train_data/ [2021/06/30 10:05:44] root INFO: label_file_list : ['/paddle/data/handwrite/train_data/5660/test.txt'] [2021/06/30 10:05:44] root INFO: name : SimpleDataSet [2021/06/30 10:05:44] root INFO: transforms : [2021/06/30 10:05:44] root INFO: DecodeImage : [2021/06/30 10:05:44] root INFO: channel_first : False [2021/06/30 10:05:44] root INFO: img_mode : BGR [2021/06/30 10:05:44] root INFO: CTCLabelEncode : None [2021/06/30 10:05:44] root INFO: RecResizeImg : [2021/06/30 10:05:44] root INFO: image_shape : [3, 32, 320] [2021/06/30 10:05:44] root INFO: KeepKeys : [2021/06/30 10:05:44] root INFO: keep_keys : ['image', 'label', 'length'] [2021/06/30 10:05:44] root INFO: loader : [2021/06/30 10:05:44] root INFO: batch_size_per_card : 512 [2021/06/30 10:05:44] root INFO: drop_last : False [2021/06/30 10:05:44] root INFO: num_workers : 8 [2021/06/30 10:05:44] root INFO: shuffle : False [2021/06/30 10:05:44] root INFO: Global : [2021/06/30 10:05:44] root INFO: cal_metric_during_train : True [2021/06/30 10:05:44] root INFO: character_dict_path : /paddle/data/handwrite/train_data/5660/5660dict.txt [2021/06/30 10:05:44] root INFO: character_type : ch [2021/06/30 10:05:44] root INFO: checkpoints : None [2021/06/30 10:05:44] root INFO: debug : False [2021/06/30 10:05:44] root INFO: distributed : False [2021/06/30 10:05:44] root INFO: epoch_num : 500 [2021/06/30 10:05:44] root INFO: eval_batch_step : [180000, 30000] [2021/06/30 10:05:44] root INFO: infer_img : doc/imgs_words/ch/word_1.jpg [2021/06/30 10:05:44] root INFO: infer_mode : False [2021/06/30 10:05:44] root INFO: log_smooth_window : 20 [2021/06/30 10:05:44] root INFO: max_text_length : 25 [2021/06/30 10:05:44] root INFO: pretrained_model : /paddle/PaddleOCR/output/rec_chinese_lite_v2.0_myhandwrite_5660_49W/best_accuracy [2021/06/30 10:05:44] root INFO: print_batch_step : 10 [2021/06/30 10:05:44] root INFO: save_epoch_step : 3 [2021/06/30 10:05:44] root INFO: save_inference_dir : None [2021/06/30 10:05:44] root INFO: save_model_dir : ./output/rec_chinese_lite_v2.0_myhandwrite_5660_500W [2021/06/30 10:05:44] root INFO: save_res_path : ./output/rec/predicts_chinese_lite_v2.0_myhandwrite_5660_500W.txt [2021/06/30 10:05:44] root INFO: use_gpu : True [2021/06/30 10:05:44] root INFO: use_space_char : False [2021/06/30 10:05:44] root INFO: use_visualdl : False [2021/06/30 10:05:44] root INFO: Loss : [2021/06/30 10:05:44] root INFO: name : CTCLoss [2021/06/30 10:05:44] root INFO: Metric : [2021/06/30 10:05:44] root INFO: main_indicator : acc [2021/06/30 10:05:44] root INFO: name : RecMetric [2021/06/30 10:05:44] root INFO: Optimizer : [2021/06/30 10:05:44] root INFO: beta1 : 0.9 [2021/06/30 10:05:44] root INFO: beta2 : 0.999 [2021/06/30 10:05:44] root INFO: lr : [2021/06/30 10:05:44] root INFO: learning_rate : 0.001 [2021/06/30 10:05:44] root INFO: name : Cosine [2021/06/30 10:05:44] root INFO: name : Adam [2021/06/30 10:05:44] root INFO: regularizer : [2021/06/30 10:05:44] root INFO: factor : 1e-05 [2021/06/30 10:05:44] root INFO: name : L2 [2021/06/30 10:05:44] root INFO: PostProcess : [2021/06/30 10:05:44] root INFO: name : CTCLabelDecode [2021/06/30 10:05:44] root INFO: Train : [2021/06/30 10:05:44] root INFO: dataset : [2021/06/30 10:05:44] root INFO: data_dir : /paddle/data/handwrite/train_data/ [2021/06/30 10:05:44] root INFO: label_file_list : ['/paddle/data/handwrite/train_data/5660/train.txt'] [2021/06/30 10:05:44] root INFO: name : SimpleDataSet [2021/06/30 10:05:44] root INFO: transforms : [2021/06/30 10:05:44] root INFO: DecodeImage : [2021/06/30 10:05:44] root INFO: channel_first : False [2021/06/30 10:05:44] root INFO: img_mode : BGR [2021/06/30 10:05:44] root INFO: RecAug : None [2021/06/30 10:05:44] root INFO: CTCLabelEncode : None [2021/06/30 10:05:44] root INFO: RecResizeImg : [2021/06/30 10:05:44] root INFO: image_shape : [3, 32, 320] [2021/06/30 10:05:44] root INFO: KeepKeys : [2021/06/30 10:05:44] root INFO: keep_keys : ['image', 'label', 'length'] [2021/06/30 10:05:44] root INFO: loader : [2021/06/30 10:05:44] root INFO: batch_size_per_card : 512 [2021/06/30 10:05:44] root INFO: drop_last : True [2021/06/30 10:05:44] root INFO: num_workers : 8 [2021/06/30 10:05:44] root INFO: shuffle : True [2021/06/30 10:05:44] root INFO: train with paddle 2.0.2 and device CUDAPlace(0) [2021/06/30 10:05:44] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/5660/train.txt'] [2021/06/30 10:06:12] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/5660/test.txt'] [2021/06/30 10:06:19] root INFO: load pretrained model from ['/paddle/PaddleOCR/output/rec_chinese_lite_v2.0_myhandwrite_5660_49W/best_accuracy'] [2021/06/30 10:06:19] root INFO: train dataloader has 9738 iters [2021/06/30 10:06:19] root INFO: valid dataloader has 989 iters [2021/06/30 10:06:19] root INFO: During the training process, after the 180000th iteration, an evaluation is run every 30000 iterations [2021/06/30 10:06:19] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/5660/train.txt'] [2021/06/30 10:10:19] root INFO: epoch: [1/500], iter: 10, lr: 0.001000, loss: 31.240917, acc: 0.056641, norm_edit_dis: 0.418718, reader_cost: 14.98379 s, batch_cost: 17.52557 s, samples: 5632, ips: 32.13591 [2021/06/30 10:11:59] root INFO: epoch: [1/500], iter: 20, lr: 0.001000, loss: 29.931261, acc: 0.057617, norm_edit_dis: 0.426505, reader_cost: 4.32411 s, batch_cost: 6.60721 s, samples: 5120, ips: 77.49108 [2021/06/30 10:13:45] root INFO: epoch: [1/500], iter: 30, lr: 0.001000, loss: 29.063553, acc: 0.059570, norm_edit_dis: 0.454388, reader_cost: 5.18631 s, batch_cost: 7.40404 s, samples: 5120, ips: 69.15142 [2021/06/30 10:16:06] root INFO: epoch: [1/500], iter: 40, lr: 0.001000, loss: 28.752716, acc: 0.068359, norm_edit_dis: 0.457807, reader_cost: 8.50194 s, batch_cost: 10.73804 s, samples: 5120, ips: 47.68094 [2021/06/30 10:18:11] root INFO: epoch: [1/500], iter: 50, lr: 0.001000, loss: 28.608841, acc: 0.066406, norm_edit_dis: 0.458741, reader_cost: 6.98738 s, batch_cost: 9.26929 s, samples: 5120, ips: 55.23617 [2021/06/30 10:19:57] root INFO: epoch: [1/500], iter: 60, lr: 0.001000, loss: 28.737865, acc: 0.067383, norm_edit_dis: 0.459099, reader_cost: 5.07168 s, batch_cost: 7.34452 s, samples: 5120, ips: 69.71185 [2021/06/30 10:21:50] root INFO: epoch: [1/500], iter: 70, lr: 0.001000, loss: 28.972126, acc: 0.064453, norm_edit_dis: 0.461046, reader_cost: 5.73199 s, batch_cost: 8.06377 s, samples: 5120, ips: 63.49388 [2021/06/30 10:23:57] root INFO: epoch: [1/500], iter: 80, lr: 0.001000, loss: 28.359682, acc: 0.065430, norm_edit_dis: 0.460278, reader_cost: 7.20473 s, batch_cost: 9.47071 s, samples: 5120, ips: 54.06142 [2021/06/30 10:26:21] root INFO: epoch: [1/500], iter: 90, lr: 0.001000, loss: 28.169189, acc: 0.069336, norm_edit_dis: 0.458578, reader_cost: 8.88785 s, batch_cost: 11.11398 s, samples: 5120, ips: 46.06812 [2021/06/30 10:28:23] root INFO: epoch: [1/500], iter: 100, lr: 0.001000, loss: 29.015282, acc: 0.068359, norm_edit_dis: 0.454898, reader_cost: 6.53351 s, batch_cost: 8.77350 s, samples: 5120, ips: 58.35754 [2021/06/30 10:30:08] root INFO: epoch: [1/500], iter: 110, lr: 0.001000, loss: 28.911865, acc: 0.065430, norm_edit_dis: 0.453865, reader_cost: 5.03723 s, batch_cost: 7.34129 s, samples: 5120, ips: 69.74254 [2021/06/30 10:31:54] root INFO: epoch: [1/500], iter: 120, lr: 0.001000, loss: 28.920834, acc: 0.063477, norm_edit_dis: 0.449316, reader_cost: 5.07936 s, batch_cost: 7.31080 s, samples: 5120, ips: 70.03335 [2021/06/30 10:34:14] root INFO: epoch: [1/500], iter: 130, lr: 0.001000, loss: 28.920834, acc: 0.065430, norm_edit_dis: 0.449223, reader_cost: 8.49405 s, batch_cost: 10.79574 s, samples: 5120, ips: 47.42610 [2021/06/30 10:36:29] root INFO: epoch: [1/500], iter: 140, lr: 0.001000, loss: 29.403336, acc: 0.063477, norm_edit_dis: 0.449998, reader_cost: 7.95247 s, batch_cost: 10.22884 s, samples: 5120, ips: 50.05454 [2021/06/30 10:38:14] root INFO: epoch: [1/500], iter: 150, lr: 0.001000, loss: 29.403336, acc: 0.062500, norm_edit_dis: 0.451379, reader_cost: 5.13431 s, batch_cost: 7.37502 s, samples: 5120, ips: 69.42358 [2021/06/30 10:40:00] root INFO: epoch: [1/500], iter: 160, lr: 0.001000, loss: 28.652584, acc: 0.065430, norm_edit_dis: 0.459963, reader_cost: 5.17615 s, batch_cost: 7.42624 s, samples: 5120, ips: 68.94468 [2021/06/30 10:42:32] root INFO: epoch: [1/500], iter: 170, lr: 0.001000, loss: 28.652584, acc: 0.063477, norm_edit_dis: 0.459191, reader_cost: 9.73111 s, batch_cost: 11.99873 s, samples: 5120, ips: 42.67117 [2021/06/30 10:44:25] root INFO: epoch: [1/500], iter: 180, lr: 0.001000, loss: 28.960621, acc: 0.065430, norm_edit_dis: 0.456993, reader_cost: 5.86574 s, batch_cost: 8.15447 s, samples: 5120, ips: 62.78764 [2021/06/30 10:46:07] root INFO: epoch: [1/500], iter: 190, lr: 0.001000, loss: 28.701799, acc: 0.066406, norm_edit_dis: 0.460723, reader_cost: 4.61426 s, batch_cost: 6.85204 s, samples: 5120, ips: 74.72232 [2021/06/30 10:48:18] root INFO: epoch: [1/500], iter: 200, lr: 0.001000, loss: 29.476135, acc: 0.066406, norm_edit_dis: 0.463056, reader_cost: 7.58683 s, batch_cost: 9.89736 s, samples: 5120, ips: 51.73096 [2021/06/30 10:50:34] root INFO: epoch: [1/500], iter: 210, lr: 0.001000, loss: 29.196236, acc: 0.067383, norm_edit_dis: 0.462320, reader_cost: 8.02476 s, batch_cost: 10.27828 s, samples: 5120, ips: 49.81378 [2021/06/30 10:52:20] root INFO: epoch: [1/500], iter: 220, lr: 0.001000, loss: 28.938604, acc: 0.063477, norm_edit_dis: 0.458144, reader_cost: 5.67016 s, batch_cost: 7.69411 s, samples: 5120, ips: 66.54444 [2021/06/30 10:54:00] root INFO: epoch: [1/500], iter: 230, lr: 0.001000, loss: 29.047894, acc: 0.062500, norm_edit_dis: 0.455604, reader_cost: 5.10638 s, batch_cost: 7.11311 s, samples: 5120, ips: 71.97981 [2021/06/30 10:56:23] root INFO: epoch: [1/500], iter: 240, lr: 0.001000, loss: 29.001678, acc: 0.064453, norm_edit_dis: 0.455803, reader_cost: 9.38144 s, batch_cost: 11.40148 s, samples: 5120, ips: 44.90644 [2021/06/30 10:58:29] root INFO: epoch: [1/500], iter: 250, lr: 0.001000, loss: 29.021385, acc: 0.064453, norm_edit_dis: 0.457864, reader_cost: 7.79869 s, batch_cost: 9.81200 s, samples: 5120, ips: 52.18100 [2021/06/30 11:00:16] root INFO: epoch: [1/500], iter: 260, lr: 0.001000, loss: 29.368792, acc: 0.064453, norm_edit_dis: 0.456196, reader_cost: 5.77732 s, batch_cost: 7.78389 s, samples: 5120, ips: 65.77686 [2021/06/30 11:01:58] root INFO: epoch: [1/500], iter: 270, lr: 0.001000, loss: 29.761681, acc: 0.058594, norm_edit_dis: 0.449271, reader_cost: 5.33553 s, batch_cost: 7.35235 s, samples: 5120, ips: 69.63758 [2021/06/30 11:04:40] root INFO: epoch: [1/500], iter: 280, lr: 0.001000, loss: 28.928726, acc: 0.058594, norm_edit_dis: 0.451067, reader_cost: 11.32355 s, batch_cost: 13.34461 s, samples: 5120, ips: 38.36756 [2021/06/30 11:06:23] root INFO: epoch: [1/500], iter: 290, lr: 0.001000, loss: 28.899677, acc: 0.062500, norm_edit_dis: 0.457633, reader_cost: 5.42592 s, batch_cost: 7.42733 s, samples: 5120, ips: 68.93462 [2021/06/30 11:08:06] root INFO: epoch: [1/500], iter: 300, lr: 0.001000, loss: 28.954216, acc: 0.064453, norm_edit_dis: 0.461166, reader_cost: 5.36566 s, batch_cost: 7.37086 s, samples: 5120, ips: 69.46268 [2021/06/30 11:09:55] root INFO: epoch: [1/500], iter: 310, lr: 0.001000, loss: 28.804794, acc: 0.066406, norm_edit_dis: 0.459190, reader_cost: 6.03951 s, batch_cost: 8.04818 s, samples: 5120, ips: 63.61684 [2021/06/30 11:12:39] root INFO: epoch: [1/500], iter: 320, lr: 0.001000, loss: 28.737762, acc: 0.064453, norm_edit_dis: 0.455933, reader_cost: 11.53702 s, batch_cost: 13.57055 s, samples: 5120, ips: 37.72876 [2021/06/30 11:14:30] root INFO: epoch: [1/500], iter: 330, lr: 0.001000, loss: 28.733232, acc: 0.065430, norm_edit_dis: 0.462456, reader_cost: 6.19043 s, batch_cost: 8.20546 s, samples: 5120, ips: 62.39749 [2021/06/30 11:16:15] root INFO: epoch: [1/500], iter: 340, lr: 0.001000, loss: 28.733232, acc: 0.067383, norm_edit_dis: 0.463902, reader_cost: 5.59001 s, batch_cost: 7.60060 s, samples: 5120, ips: 67.36306 [2021/06/30 11:17:57] root INFO: epoch: [1/500], iter: 350, lr: 0.001000, loss: 28.968098, acc: 0.062500, norm_edit_dis: 0.449038, reader_cost: 5.36341 s, batch_cost: 7.37285 s, samples: 5120, ips: 69.44393 [2021/06/30 11:20:21] root INFO: epoch: [1/500], iter: 360, lr: 0.001000, loss: 29.146383, acc: 0.063477, norm_edit_dis: 0.450650, reader_cost: 9.52700 s, batch_cost: 11.55431 s, samples: 5120, ips: 44.31247 [2021/06/30 11:22:22] root INFO: epoch: [1/500], iter: 370, lr: 0.001000, loss: 29.051979, acc: 0.066406, norm_edit_dis: 0.450803, reader_cost: 7.20553 s, batch_cost: 9.21864 s, samples: 5120, ips: 55.53965 [2021/06/30 11:24:12] root INFO: epoch: [1/500], iter: 380, lr: 0.001000, loss: 28.745113, acc: 0.061523, norm_edit_dis: 0.450217, reader_cost: 6.10408 s, batch_cost: 8.12954 s, samples: 5120, ips: 62.98018 [2021/06/30 11:25:58] root INFO: epoch: [1/500], iter: 390, lr: 0.001000, loss: 29.158270, acc: 0.060547, norm_edit_dis: 0.456435, reader_cost: 5.68604 s, batch_cost: 7.71055 s, samples: 5120, ips: 66.40254 [2021/06/30 11:28:24] root INFO: epoch: [1/500], iter: 400, lr: 0.001000, loss: 29.241486, acc: 0.062500, norm_edit_dis: 0.459944, reader_cost: 9.71734 s, batch_cost: 11.73385 s, samples: 5120, ips: 43.63445 [2021/06/30 11:30:33] root INFO: epoch: [1/500], iter: 410, lr: 0.001000, loss: 29.073112, acc: 0.063477, norm_edit_dis: 0.461218, reader_cost: 8.07536 s, batch_cost: 10.09336 s, samples: 5120, ips: 50.72641 [2021/06/30 11:32:24] root INFO: epoch: [1/500], iter: 420, lr: 0.001000, loss: 29.064835, acc: 0.070312, norm_edit_dis: 0.461727, reader_cost: 6.17176 s, batch_cost: 8.18808 s, samples: 5120, ips: 62.52995 [2021/06/30 11:34:10] root INFO: epoch: [1/500], iter: 430, lr: 0.001000, loss: 29.064835, acc: 0.070312, norm_edit_dis: 0.464027, reader_cost: 5.75984 s, batch_cost: 7.77313 s, samples: 5120, ips: 65.86797 [2021/06/30 11:36:21] root INFO: epoch: [1/500], iter: 440, lr: 0.001000, loss: 28.942097, acc: 0.066406, norm_edit_dis: 0.460420, reader_cost: 8.17048 s, batch_cost: 10.17095 s, samples: 5120, ips: 50.33945 [2021/06/30 11:38:51] root INFO: epoch: [1/500], iter: 450, lr: 0.001000, loss: 28.776918, acc: 0.066406, norm_edit_dis: 0.459700, reader_cost: 10.11811 s, batch_cost: 12.11821 s, samples: 5120, ips: 42.25046 [2021/06/30 11:40:31] root INFO: epoch: [1/500], iter: 460, lr: 0.001000, loss: 28.640152, acc: 0.066406, norm_edit_dis: 0.464868, reader_cost: 5.17719 s, batch_cost: 7.19007 s, samples: 5120, ips: 71.20930 [2021/06/30 11:42:16] root INFO: epoch: [1/500], iter: 470, lr: 0.001000, loss: 28.173771, acc: 0.067383, norm_edit_dis: 0.464868, reader_cost: 5.62408 s, batch_cost: 7.63949 s, samples: 5120, ips: 67.02021 [2021/06/30 11:44:29] root INFO: epoch: [1/500], iter: 480, lr: 0.001000, loss: 28.759075, acc: 0.067383, norm_edit_dis: 0.464287, reader_cost: 8.42004 s, batch_cost: 10.42875 s, samples: 5120, ips: 49.09505 [2021/06/30 11:46:47] root INFO: epoch: [1/500], iter: 490, lr: 0.001000, loss: 28.696434, acc: 0.070312, norm_edit_dis: 0.466792, reader_cost: 8.84879 s, batch_cost: 10.86693 s, samples: 5120, ips: 47.11541 [2021/06/30 11:48:34] root INFO: epoch: [1/500], iter: 500, lr: 0.001000, loss: 28.362125, acc: 0.070312, norm_edit_dis: 0.472159, reader_cost: 5.78696 s, batch_cost: 7.79431 s, samples: 5120, ips: 65.68893 [2021/06/30 11:50:17] root INFO: epoch: [1/500], iter: 510, lr: 0.001000, loss: 28.226355, acc: 0.066406, norm_edit_dis: 0.473989, reader_cost: 5.42195 s, batch_cost: 7.42955 s, samples: 5120, ips: 68.91397 [2021/06/30 11:52:28] root INFO: epoch: [1/500], iter: 520, lr: 0.001000, loss: 28.292847, acc: 0.066406, norm_edit_dis: 0.470583, reader_cost: 8.22645 s, batch_cost: 10.24074 s, samples: 5120, ips: 49.99641 [2021/06/30 11:54:40] root INFO: epoch: [1/500], iter: 530, lr: 0.001000, loss: 28.736652, acc: 0.071289, norm_edit_dis: 0.468598, reader_cost: 8.24087 s, batch_cost: 10.27221 s, samples: 5120, ips: 49.84320 [2021/06/30 11:56:20] root INFO: epoch: [1/500], iter: 540, lr: 0.001000, loss: 28.425404, acc: 0.072266, norm_edit_dis: 0.473481, reader_cost: 5.15648 s, batch_cost: 7.16887 s, samples: 5120, ips: 71.41987 [2021/06/30 11:58:05] root INFO: epoch: [1/500], iter: 550, lr: 0.001000, loss: 28.159788, acc: 0.072266, norm_edit_dis: 0.471328, reader_cost: 5.54386 s, batch_cost: 7.55497 s, samples: 5120, ips: 67.76997 [2021/06/30 12:00:35] root INFO: epoch: [1/500], iter: 560, lr: 0.001000, loss: 28.237759, acc: 0.076172, norm_edit_dis: 0.471328, reader_cost: 10.07634 s, batch_cost: 12.09368 s, samples: 5120, ips: 42.33615 [2021/06/30 12:02:34] root INFO: epoch: [1/500], iter: 570, lr: 0.001000, loss: 28.269911, acc: 0.074219, norm_edit_dis: 0.467947, reader_cost: 7.06222 s, batch_cost: 9.07959 s, samples: 5120, ips: 56.39021 [2021/06/30 12:04:22] root INFO: epoch: [1/500], iter: 580, lr: 0.001000, loss: 28.190353, acc: 0.066406, norm_edit_dis: 0.467210, reader_cost: 5.86674 s, batch_cost: 7.88063 s, samples: 5120, ips: 64.96946 [2021/06/30 12:06:09] root INFO: epoch: [1/500], iter: 590, lr: 0.001000, loss: 27.891895, acc: 0.065430, norm_edit_dis: 0.470969, reader_cost: 5.81250 s, batch_cost: 7.82621 s, samples: 5120, ips: 65.42120 [2021/06/30 12:08:51] root INFO: epoch: [1/500], iter: 600, lr: 0.001000, loss: 27.933929, acc: 0.066406, norm_edit_dis: 0.477819, reader_cost: 11.27344 s, batch_cost: 13.28886 s, samples: 5120, ips: 38.52850 [2021/06/30 12:10:48] root INFO: epoch: [1/500], iter: 610, lr: 0.001000, loss: 28.078323, acc: 0.072266, norm_edit_dis: 0.474948, reader_cost: 6.84774 s, batch_cost: 8.84874 s, samples: 5120, ips: 57.86134 [2021/06/30 12:12:30] root INFO: epoch: [1/500], iter: 620, lr: 0.001000, loss: 28.703627, acc: 0.072266, norm_edit_dis: 0.466865, reader_cost: 5.39026 s, batch_cost: 7.39218 s, samples: 5120, ips: 69.26243 [2021/06/30 12:14:18] root INFO: epoch: [1/500], iter: 630, lr: 0.001000, loss: 28.899885, acc: 0.069336, norm_edit_dis: 0.459304, reader_cost: 5.94170 s, batch_cost: 7.95324 s, samples: 5120, ips: 64.37625 [2021/06/30 12:16:56] root INFO: epoch: [1/500], iter: 640, lr: 0.001000, loss: 28.561722, acc: 0.067383, norm_edit_dis: 0.461272, reader_cost: 10.87176 s, batch_cost: 12.88228 s, samples: 5120, ips: 39.74452 [2021/06/30 12:18:51] root INFO: epoch: [1/500], iter: 650, lr: 0.001000, loss: 28.615982, acc: 0.065430, norm_edit_dis: 0.461272, reader_cost: 6.67661 s, batch_cost: 8.68191 s, samples: 5120, ips: 58.97323 [2021/06/30 12:20:39] root INFO: epoch: [1/500], iter: 660, lr: 0.001000, loss: 28.805407, acc: 0.064453, norm_edit_dis: 0.460906, reader_cost: 5.85671 s, batch_cost: 7.87161 s, samples: 5120, ips: 65.04387 [2021/06/30 12:22:28] root INFO: epoch: [1/500], iter: 670, lr: 0.001000, loss: 28.465267, acc: 0.069336, norm_edit_dis: 0.470795, reader_cost: 6.01108 s, batch_cost: 8.02061 s, samples: 5120, ips: 63.83558 [2021/06/30 12:24:47] root INFO: epoch: [1/500], iter: 680, lr: 0.001000, loss: 28.128532, acc: 0.067383, norm_edit_dis: 0.468589, reader_cost: 8.96983 s, batch_cost: 10.99364 s, samples: 5120, ips: 46.57236 [2021/06/30 12:26:40] root INFO: epoch: [1/500], iter: 690, lr: 0.001000, loss: 28.108297, acc: 0.064453, norm_edit_dis: 0.461884, reader_cost: 6.44312 s, batch_cost: 8.46104 s, samples: 5120, ips: 60.51264 [2021/06/30 12:28:22] root INFO: epoch: [1/500], iter: 700, lr: 0.001000, loss: 28.473877, acc: 0.063477, norm_edit_dis: 0.463387, reader_cost: 5.25148 s, batch_cost: 7.26671 s, samples: 5120, ips: 70.45830 [2021/06/30 12:30:13] root INFO: epoch: [1/500], iter: 710, lr: 0.001000, loss: 28.715656, acc: 0.065430, norm_edit_dis: 0.464576, reader_cost: 6.30455 s, batch_cost: 8.31207 s, samples: 5120, ips: 61.59716 [2021/06/30 12:32:33] root INFO: epoch: [1/500], iter: 720, lr: 0.001000, loss: 28.714520, acc: 0.068359, norm_edit_dis: 0.468680, reader_cost: 9.09563 s, batch_cost: 11.10394 s, samples: 5120, ips: 46.10974 [2021/06/30 12:34:52] root INFO: epoch: [1/500], iter: 730, lr: 0.001000, loss: 28.349289, acc: 0.070312, norm_edit_dis: 0.468098, reader_cost: 9.05203 s, batch_cost: 11.05918 s, samples: 5120, ips: 46.29636 [2021/06/30 12:36:31] root INFO: epoch: [1/500], iter: 740, lr: 0.001000, loss: 28.157452, acc: 0.068359, norm_edit_dis: 0.466138, reader_cost: 5.03931 s, batch_cost: 7.04134 s, samples: 5120, ips: 72.71340 [2021/06/30 12:38:15] root INFO: epoch: [1/500], iter: 750, lr: 0.001000, loss: 28.477364, acc: 0.066406, norm_edit_dis: 0.466440, reader_cost: 5.44301 s, batch_cost: 7.45433 s, samples: 5120, ips: 68.68490 [2021/06/30 12:40:25] root INFO: epoch: [1/500], iter: 760, lr: 0.001000, loss: 28.604227, acc: 0.066406, norm_edit_dis: 0.468272, reader_cost: 8.13835 s, batch_cost: 10.15402 s, samples: 5120, ips: 50.42339 [2021/06/30 12:42:48] root INFO: epoch: [1/500], iter: 770, lr: 0.001000, loss: 28.356033, acc: 0.067383, norm_edit_dis: 0.469472, reader_cost: 9.39149 s, batch_cost: 11.40433 s, samples: 5120, ips: 44.89522 [2021/06/30 12:44:31] root INFO: epoch: [1/500], iter: 780, lr: 0.001000, loss: 28.179569, acc: 0.068359, norm_edit_dis: 0.468194, reader_cost: 5.44025 s, batch_cost: 7.44190 s, samples: 5120, ips: 68.79964 [2021/06/30 12:46:19] root INFO: epoch: [1/500], iter: 790, lr: 0.001000, loss: 28.387901, acc: 0.069336, norm_edit_dis: 0.468575, reader_cost: 5.89421 s, batch_cost: 7.91266 s, samples: 5120, ips: 64.70645 [2021/06/30 12:48:28] root INFO: epoch: [1/500], iter: 800, lr: 0.001000, loss: 28.387901, acc: 0.074219, norm_edit_dis: 0.470769, reader_cost: 7.97965 s, batch_cost: 9.99731 s, samples: 5120, ips: 51.21377 [2021/06/30 12:50:35] root INFO: epoch: [1/500], iter: 810, lr: 0.001000, loss: 27.823750, acc: 0.076172, norm_edit_dis: 0.471394, reader_cost: 7.84990 s, batch_cost: 9.85493 s, samples: 5120, ips: 51.95371 [2021/06/30 12:52:26] root INFO: epoch: [1/500], iter: 820, lr: 0.001000, loss: 28.115864, acc: 0.073242, norm_edit_dis: 0.470556, reader_cost: 6.19001 s, batch_cost: 8.21699 s, samples: 5120, ips: 62.30991 [2021/06/30 12:54:09] root INFO: epoch: [1/500], iter: 830, lr: 0.001000, loss: 28.710503, acc: 0.072266, norm_edit_dis: 0.469747, reader_cost: 5.46389 s, batch_cost: 7.47894 s, samples: 5120, ips: 68.45886 [2021/06/30 12:56:23] root INFO: epoch: [1/500], iter: 840, lr: 0.001000, loss: 29.178604, acc: 0.073242, norm_edit_dis: 0.467639, reader_cost: 8.52222 s, batch_cost: 10.52756 s, samples: 5120, ips: 48.63423 [2021/06/30 12:58:39] root INFO: epoch: [1/500], iter: 850, lr: 0.001000, loss: 28.458214, acc: 0.070312, norm_edit_dis: 0.470220, reader_cost: 8.67604 s, batch_cost: 10.69119 s, samples: 5120, ips: 47.88989 [2021/06/30 13:00:27] root INFO: epoch: [1/500], iter: 860, lr: 0.001000, loss: 27.963924, acc: 0.068359, norm_edit_dis: 0.470177, reader_cost: 5.88580 s, batch_cost: 7.91773 s, samples: 5120, ips: 64.66500 [2021/06/30 13:02:13] root INFO: epoch: [1/500], iter: 870, lr: 0.001000, loss: 29.029671, acc: 0.073242, norm_edit_dis: 0.466783, reader_cost: 5.70341 s, batch_cost: 7.71282 s, samples: 5120, ips: 66.38298 [2021/06/30 13:04:19] root INFO: epoch: [1/500], iter: 880, lr: 0.001000, loss: 28.463211, acc: 0.068359, norm_edit_dis: 0.464594, reader_cost: 7.62302 s, batch_cost: 9.63667 s, samples: 5120, ips: 53.13039 [2021/06/30 13:06:29] root INFO: epoch: [1/500], iter: 890, lr: 0.001000, loss: 28.166525, acc: 0.068359, norm_edit_dis: 0.466103, reader_cost: 8.12645 s, batch_cost: 10.13842 s, samples: 5120, ips: 50.50098 [2021/06/30 13:08:10] root INFO: epoch: [1/500], iter: 900, lr: 0.001000, loss: 28.250711, acc: 0.069336, norm_edit_dis: 0.465625, reader_cost: 5.14691 s, batch_cost: 7.16375 s, samples: 5120, ips: 71.47092 [2021/06/30 13:09:53] root INFO: epoch: [1/500], iter: 910, lr: 0.001000, loss: 28.285503, acc: 0.063477, norm_edit_dis: 0.465499, reader_cost: 5.39467 s, batch_cost: 7.40608 s, samples: 5120, ips: 69.13239 [2021/06/30 13:12:14] root INFO: epoch: [1/500], iter: 920, lr: 0.001000, loss: 28.497967, acc: 0.068359, norm_edit_dis: 0.468391, reader_cost: 9.28608 s, batch_cost: 11.30375 s, samples: 5120, ips: 45.29472 [2021/06/30 13:14:20] root INFO: epoch: [1/500], iter: 930, lr: 0.001000, loss: 28.180235, acc: 0.071289, norm_edit_dis: 0.470724, reader_cost: 7.59703 s, batch_cost: 9.63173 s, samples: 5120, ips: 53.15763 [2021/06/30 13:16:04] root INFO: epoch: [1/500], iter: 940, lr: 0.001000, loss: 27.983078, acc: 0.072266, norm_edit_dis: 0.466667, reader_cost: 5.58320 s, batch_cost: 7.60704 s, samples: 5120, ips: 67.30606 [2021/06/30 13:17:49] root INFO: epoch: [1/500], iter: 950, lr: 0.001000, loss: 27.746002, acc: 0.070312, norm_edit_dis: 0.466667, reader_cost: 5.60527 s, batch_cost: 7.61596 s, samples: 5120, ips: 67.22727 [2021/06/30 13:20:21] root INFO: epoch: [1/500], iter: 960, lr: 0.001000, loss: 28.008808, acc: 0.070312, norm_edit_dis: 0.471570, reader_cost: 10.29654 s, batch_cost: 12.31794 s, samples: 5120, ips: 41.56539 [2021/06/30 13:22:15] root INFO: epoch: [1/500], iter: 970, lr: 0.001000, loss: 28.126303, acc: 0.072266, norm_edit_dis: 0.473188, reader_cost: 6.43201 s, batch_cost: 8.45166 s, samples: 5120, ips: 60.57979 [2021/06/30 13:24:02] root INFO: epoch: [1/500], iter: 980, lr: 0.001000, loss: 28.711082, acc: 0.069336, norm_edit_dis: 0.468217, reader_cost: 5.77872 s, batch_cost: 7.79508 s, samples: 5120, ips: 65.68242 [2021/06/30 13:25:42] root INFO: epoch: [1/500], iter: 990, lr: 0.001000, loss: 28.707054, acc: 0.066406, norm_edit_dis: 0.460275, reader_cost: 5.10476 s, batch_cost: 7.12058 s, samples: 5120, ips: 71.90430 [2021/06/30 13:28:16] root INFO: epoch: [1/500], iter: 1000, lr: 0.001000, loss: 28.655977, acc: 0.063477, norm_edit_dis: 0.460463, reader_cost: 10.54009 s, batch_cost: 12.55472 s, samples: 5120, ips: 40.78149 [2021/06/30 13:29:54] root INFO: epoch: [1/500], iter: 1010, lr: 0.001000, loss: 28.081484, acc: 0.066406, norm_edit_dis: 0.463683, reader_cost: 4.90912 s, batch_cost: 6.93585 s, samples: 5120, ips: 73.81941 [2021/06/30 13:31:42] root INFO: epoch: [1/500], iter: 1020, lr: 0.001000, loss: 28.431808, acc: 0.066406, norm_edit_dis: 0.467774, reader_cost: 5.87391 s, batch_cost: 7.87105 s, samples: 5120, ips: 65.04853 [2021/06/30 13:33:25] root INFO: epoch: [1/500], iter: 1030, lr: 0.001000, loss: 28.702068, acc: 0.066406, norm_edit_dis: 0.466427, reader_cost: 5.50698 s, batch_cost: 7.51992 s, samples: 5120, ips: 68.08580 [2021/06/30 13:35:48] root INFO: epoch: [1/500], iter: 1040, lr: 0.001000, loss: 28.051796, acc: 0.068359, norm_edit_dis: 0.467071, reader_cost: 9.37153 s, batch_cost: 11.38322 s, samples: 5120, ips: 44.97849 [2021/06/30 13:38:00] root INFO: epoch: [1/500], iter: 1050, lr: 0.001000, loss: 27.695133, acc: 0.072266, norm_edit_dis: 0.469461, reader_cost: 8.25736 s, batch_cost: 10.26917 s, samples: 5120, ips: 49.85796 [2021/06/30 13:39:46] root INFO: epoch: [1/500], iter: 1060, lr: 0.001000, loss: 27.637615, acc: 0.072266, norm_edit_dis: 0.477125, reader_cost: 5.73387 s, batch_cost: 7.73790 s, samples: 5120, ips: 66.16784 [2021/06/30 13:41:32] root INFO: epoch: [1/500], iter: 1070, lr: 0.001000, loss: 28.487730, acc: 0.070312, norm_edit_dis: 0.474837, reader_cost: 5.65326 s, batch_cost: 7.67167 s, samples: 5120, ips: 66.73904 [2021/06/30 13:43:31] root INFO: epoch: [1/500], iter: 1080, lr: 0.001000, loss: 28.685019, acc: 0.069336, norm_edit_dis: 0.472277, reader_cost: 7.01997 s, batch_cost: 9.02523 s, samples: 5120, ips: 56.72987 [2021/06/30 13:46:16] root INFO: epoch: [1/500], iter: 1090, lr: 0.001000, loss: 28.638580, acc: 0.072266, norm_edit_dis: 0.475241, reader_cost: 11.71552 s, batch_cost: 13.70732 s, samples: 5120, ips: 37.35229 [2021/06/30 13:48:01] root INFO: epoch: [1/500], iter: 1100, lr: 0.001000, loss: 28.413712, acc: 0.069336, norm_edit_dis: 0.470005, reader_cost: 5.62880 s, batch_cost: 7.64920 s, samples: 5120, ips: 66.93507 [2021/06/30 13:49:38] root INFO: epoch: [1/500], iter: 1110, lr: 0.001000, loss: 28.413712, acc: 0.070312, norm_edit_dis: 0.466420, reader_cost: 4.80654 s, batch_cost: 6.81625 s, samples: 5120, ips: 75.11465 [2021/06/30 13:51:40] root INFO: epoch: [1/500], iter: 1120, lr: 0.001000, loss: 28.347948, acc: 0.069336, norm_edit_dis: 0.472470, reader_cost: 7.23361 s, batch_cost: 9.25127 s, samples: 5120, ips: 55.34376 [2021/06/30 13:54:09] root INFO: epoch: [1/500], iter: 1130, lr: 0.001000, loss: 28.396919, acc: 0.069336, norm_edit_dis: 0.468906, reader_cost: 10.00217 s, batch_cost: 12.03964 s, samples: 5120, ips: 42.52620 [2021/06/30 13:55:53] root INFO: epoch: [1/500], iter: 1140, lr: 0.001000, loss: 29.213041, acc: 0.070312, norm_edit_dis: 0.453394, reader_cost: 5.57170 s, batch_cost: 7.57531 s, samples: 5120, ips: 67.58799 [2021/06/30 13:57:43] root INFO: epoch: [1/500], iter: 1150, lr: 0.001000, loss: 28.723248, acc: 0.065430, norm_edit_dis: 0.458687, reader_cost: 6.08701 s, batch_cost: 8.09834 s, samples: 5120, ips: 63.22282 [2021/06/30 13:59:42] root INFO: epoch: [1/500], iter: 1160, lr: 0.001000, loss: 28.491301, acc: 0.066406, norm_edit_dis: 0.468724, reader_cost: 7.07698 s, batch_cost: 9.08576 s, samples: 5120, ips: 56.35192 [

Lanme commented 3 years ago

请问跑的结果如何？

Evezerest commented 3 years ago

可以参考一下通用模型的训练时长：500W数据，4卡V100，共训练6天

paddle-bot-old[bot] commented 2 years ago

Since you haven\'t replied for more than 3 months, we have closed this issue/pr. If the problem is not solved or there is a follow-up one, please reopen it at any time and we will continue to follow up. It is recommended to pull and try the latest code first. 由于您超过三个月未回复，我们将关闭这个issue/pr。 若问题未解决或有后续问题，请随时重新打开（建议先拉取最新代码进行尝试），我们会继续跟进。

yaofanji commented 2 years ago

@littletomatodonkey 大佬现在我准备了500W左右的数据 在之前陆陆续续训练的模型基础之上训练 显卡v100 32G显存 我估计了一下一轮大概30多小时 这训练是要按照年计算吗.........请问大佬 单卡情况下有加快训练的办法吗 [2021/06/30 10:05:44] root INFO: Architecture : [2021/06/30 10:05:44] root INFO: Backbone : [2021/06/30 10:05:44] root INFO: model_name : small [2021/06/30 10:05:44] root INFO: name : MobileNetV3 [2021/06/30 10:05:44] root INFO: scale : 0.5 [2021/06/30 10:05:44] root INFO: small_stride : [1, 2, 2, 2] [2021/06/30 10:05:44] root INFO: Head : [2021/06/30 10:05:44] root INFO: fc_decay : 1e-05 [2021/06/30 10:05:44] root INFO: name : CTCHead [2021/06/30 10:05:44] root INFO: Neck : [2021/06/30 10:05:44] root INFO: encoder_type : rnn [2021/06/30 10:05:44] root INFO: hidden_size : 48 [2021/06/30 10:05:44] root INFO: name : SequenceEncoder [2021/06/30 10:05:44] root INFO: Transform : None [2021/06/30 10:05:44] root INFO: algorithm : CRNN [2021/06/30 10:05:44] root INFO: model_type : rec [2021/06/30 10:05:44] root INFO: Eval : [2021/06/30 10:05:44] root INFO: dataset : [2021/06/30 10:05:44] root INFO: data_dir : /paddle/data/handwrite/train_data/ [2021/06/30 10:05:44] root INFO: label_file_list : ['/paddle/data/handwrite/train_data/5660/test.txt'] [2021/06/30 10:05:44] root INFO: name : SimpleDataSet [2021/06/30 10:05:44] root INFO: transforms : [2021/06/30 10:05:44] root INFO: DecodeImage : [2021/06/30 10:05:44] root INFO: channel_first : False [2021/06/30 10:05:44] root INFO: img_mode : BGR [2021/06/30 10:05:44] root INFO: CTCLabelEncode : None [2021/06/30 10:05:44] root INFO: RecResizeImg : [2021/06/30 10:05:44] root INFO: image_shape : [3, 32, 320] [2021/06/30 10:05:44] root INFO: KeepKeys : [2021/06/30 10:05:44] root INFO: keep_keys : ['image', 'label', 'length'] [2021/06/30 10:05:44] root INFO: loader : [2021/06/30 10:05:44] root INFO: batch_size_per_card : 512 [2021/06/30 10:05:44] root INFO: drop_last : False [2021/06/30 10:05:44] root INFO: num_workers : 8 [2021/06/30 10:05:44] root INFO: shuffle : False [2021/06/30 10:05:44] root INFO: Global : [2021/06/30 10:05:44] root INFO: cal_metric_during_train : True [2021/06/30 10:05:44] root INFO: character_dict_path : /paddle/data/handwrite/train_data/5660/5660dict.txt [2021/06/30 10:05:44] root INFO: character_type : ch [2021/06/30 10:05:44] root INFO: checkpoints : None [2021/06/30 10:05:44] root INFO: debug : False [2021/06/30 10:05:44] root INFO: distributed : False [2021/06/30 10:05:44] root INFO: epoch_num : 500 [2021/06/30 10:05:44] root INFO: eval_batch_step : [180000, 30000] [2021/06/30 10:05:44] root INFO: infer_img : doc/imgs_words/ch/word_1.jpg [2021/06/30 10:05:44] root INFO: infer_mode : False [2021/06/30 10:05:44] root INFO: log_smooth_window : 20 [2021/06/30 10:05:44] root INFO: max_text_length : 25 [2021/06/30 10:05:44] root INFO: pretrained_model : /paddle/PaddleOCR/output/rec_chinese_lite_v2.0_myhandwrite_5660_49W/best_accuracy [2021/06/30 10:05:44] root INFO: print_batch_step : 10 [2021/06/30 10:05:44] root INFO: save_epoch_step : 3 [2021/06/30 10:05:44] root INFO: save_inference_dir : None [2021/06/30 10:05:44] root INFO: save_model_dir : ./output/rec_chinese_lite_v2.0_myhandwrite_5660_500W [2021/06/30 10:05:44] root INFO: save_res_path : ./output/rec/predicts_chinese_lite_v2.0_myhandwrite_5660_500W.txt [2021/06/30 10:05:44] root INFO: use_gpu : True [2021/06/30 10:05:44] root INFO: use_space_char : False [2021/06/30 10:05:44] root INFO: use_visualdl : False [2021/06/30 10:05:44] root INFO: Loss : [2021/06/30 10:05:44] root INFO: name : CTCLoss [2021/06/30 10:05:44] root INFO: Metric : [2021/06/30 10:05:44] root INFO: main_indicator : acc [2021/06/30 10:05:44] root INFO: name : RecMetric [2021/06/30 10:05:44] root INFO: Optimizer : [2021/06/30 10:05:44] root INFO: beta1 : 0.9 [2021/06/30 10:05:44] root INFO: beta2 : 0.999 [2021/06/30 10:05:44] root INFO: lr : [2021/06/30 10:05:44] root INFO: learning_rate : 0.001 [2021/06/30 10:05:44] root INFO: name : Cosine [2021/06/30 10:05:44] root INFO: name : Adam [2021/06/30 10:05:44] root INFO: regularizer : [2021/06/30 10:05:44] root INFO: factor : 1e-05 [2021/06/30 10:05:44] root INFO: name : L2 [2021/06/30 10:05:44] root INFO: PostProcess : [2021/06/30 10:05:44] root INFO: name : CTCLabelDecode [2021/06/30 10:05:44] root INFO: Train : [2021/06/30 10:05:44] root INFO: dataset : [2021/06/30 10:05:44] root INFO: data_dir : /paddle/data/handwrite/train_data/ [2021/06/30 10:05:44] root INFO: label_file_list : ['/paddle/data/handwrite/train_data/5660/train.txt'] [2021/06/30 10:05:44] root INFO: name : SimpleDataSet [2021/06/30 10:05:44] root INFO: transforms : [2021/06/30 10:05:44] root INFO: DecodeImage : [2021/06/30 10:05:44] root INFO: channel_first : False [2021/06/30 10:05:44] root INFO: img_mode : BGR [2021/06/30 10:05:44] root INFO: RecAug : None [2021/06/30 10:05:44] root INFO: CTCLabelEncode : None [2021/06/30 10:05:44] root INFO: RecResizeImg : [2021/06/30 10:05:44] root INFO: image_shape : [3, 32, 320] [2021/06/30 10:05:44] root INFO: KeepKeys : [2021/06/30 10:05:44] root INFO: keep_keys : ['image', 'label', 'length'] [2021/06/30 10:05:44] root INFO: loader : [2021/06/30 10:05:44] root INFO: batch_size_per_card : 512 [2021/06/30 10:05:44] root INFO: drop_last : True [2021/06/30 10:05:44] root INFO: num_workers : 8 [2021/06/30 10:05:44] root INFO: shuffle : True [2021/06/30 10:05:44] root INFO: train with paddle 2.0.2 and device CUDAPlace(0) [2021/06/30 10:05:44] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/5660/train.txt'] [2021/06/30 10:06:12] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/5660/test.txt'] [2021/06/30 10:06:19] root INFO: load pretrained model from ['/paddle/PaddleOCR/output/rec_chinese_lite_v2.0_myhandwrite_5660_49W/best_accuracy'] [2021/06/30 10:06:19] root INFO: train dataloader has 9738 iters [2021/06/30 10:06:19] root INFO: valid dataloader has 989 iters [2021/06/30 10:06:19] root INFO: During the training process, after the 180000th iteration, an evaluation is run every 30000 iterations [2021/06/30 10:06:19] root INFO: Initialize indexs of datasets:['/paddle/data/handwrite/train_data/5660/train.txt'] [2021/06/30 10:10:19] root INFO: epoch: [1/500], iter: 10, lr: 0.001000, loss: 31.240917, acc: 0.056641, norm_edit_dis: 0.418718, reader_cost: 14.98379 s, batch_cost: 17.52557 s, samples: 5632, ips: 32.13591 [2021/06/30 10:11:59] root INFO: epoch: [1/500], iter: 20, lr: 0.001000, loss: 29.931261, acc: 0.057617, norm_edit_dis: 0.426505, reader_cost: 4.32411 s, batch_cost: 6.60721 s, samples: 5120, ips: 77.49108 [2021/06/30 10:13:45] root INFO: epoch: [1/500], iter: 30, lr: 0.001000, loss: 29.063553, acc: 0.059570, norm_edit_dis: 0.454388, reader_cost: 5.18631 s, batch_cost: 7.40404 s, samples: 5120, ips: 69.15142 [2021/06/30 10:16:06] root INFO: epoch: [1/500], iter: 40, lr: 0.001000, loss: 28.752716, acc: 0.068359, norm_edit_dis: 0.457807, reader_cost: 8.50194 s, batch_cost: 10.73804 s, samples: 5120, ips: 47.68094 [2021/06/30 10:18:11] root INFO: epoch: [1/500], iter: 50, lr: 0.001000, loss: 28.608841, acc: 0.066406, norm_edit_dis: 0.458741, reader_cost: 6.98738 s, batch_cost: 9.26929 s, samples: 5120, ips: 55.23617 [2021/06/30 10:19:57] root INFO: epoch: [1/500], iter: 60, lr: 0.001000, loss: 28.737865, acc: 0.067383, norm_edit_dis: 0.459099, reader_cost: 5.07168 s, batch_cost: 7.34452 s, samples: 5120, ips: 69.71185 [2021/06/30 10:21:50] root INFO: epoch: [1/500], iter: 70, lr: 0.001000, loss: 28.972126, acc: 0.064453, norm_edit_dis: 0.461046, reader_cost: 5.73199 s, batch_cost: 8.06377 s, samples: 5120, ips: 63.49388 [2021/06/30 10:23:57] root INFO: epoch: [1/500], iter: 80, lr: 0.001000, loss: 28.359682, acc: 0.065430, norm_edit_dis: 0.460278, reader_cost: 7.20473 s, batch_cost: 9.47071 s, samples: 5120, ips: 54.06142 [2021/06/30 10:26:21] root INFO: epoch: [1/500], iter: 90, lr: 0.001000, loss: 28.169189, acc: 0.069336, norm_edit_dis: 0.458578, reader_cost: 8.88785 s, batch_cost: 11.11398 s, samples: 5120, ips: 46.06812 [2021/06/30 10:28:23] root INFO: epoch: [1/500], iter: 100, lr: 0.001000, loss: 29.015282, acc: 0.068359, norm_edit_dis: 0.454898, reader_cost: 6.53351 s, batch_cost: 8.77350 s, samples: 5120, ips: 58.35754 [2021/06/30 10:30:08] root INFO: epoch: [1/500], iter: 110, lr: 0.001000, loss: 28.911865, acc: 0.065430, norm_edit_dis: 0.453865, reader_cost: 5.03723 s, batch_cost: 7.34129 s, samples: 5120, ips: 69.74254 [2021/06/30 10:31:54] root INFO: epoch: [1/500], iter: 120, lr: 0.001000, loss: 28.920834, acc: 0.063477, norm_edit_dis: 0.449316, reader_cost: 5.07936 s, batch_cost: 7.31080 s, samples: 5120, ips: 70.03335 [2021/06/30 10:34:14] root INFO: epoch: [1/500], iter: 130, lr: 0.001000, loss: 28.920834, acc: 0.065430, norm_edit_dis: 0.449223, reader_cost: 8.49405 s, batch_cost: 10.79574 s, samples: 5120, ips: 47.42610 [2021/06/30 10:36:29] root INFO: epoch: [1/500], iter: 140, lr: 0.001000, loss: 29.403336, acc: 0.063477, norm_edit_dis: 0.449998, reader_cost: 7.95247 s, batch_cost: 10.22884 s, samples: 5120, ips: 50.05454 [2021/06/30 10:38:14] root INFO: epoch: [1/500], iter: 150, lr: 0.001000, loss: 29.403336, acc: 0.062500, norm_edit_dis: 0.451379, reader_cost: 5.13431 s, batch_cost: 7.37502 s, samples: 5120, ips: 69.42358 [2021/06/30 10:40:00] root INFO: epoch: [1/500], iter: 160, lr: 0.001000, loss: 28.652584, acc: 0.065430, norm_edit_dis: 0.459963, reader_cost: 5.17615 s, batch_cost: 7.42624 s, samples: 5120, ips: 68.94468 [2021/06/30 10:42:32] root INFO: epoch: [1/500], iter: 170, lr: 0.001000, loss: 28.652584, acc: 0.063477, norm_edit_dis: 0.459191, reader_cost: 9.73111 s, batch_cost: 11.99873 s, samples: 5120, ips: 42.67117 [2021/06/30 10:44:25] root INFO: epoch: [1/500], iter: 180, lr: 0.001000, loss: 28.960621, acc: 0.065430, norm_edit_dis: 0.456993, reader_cost: 5.86574 s, batch_cost: 8.15447 s, samples: 5120, ips: 62.78764 [2021/06/30 10:46:07] root INFO: epoch: [1/500], iter: 190, lr: 0.001000, loss: 28.701799, acc: 0.066406, norm_edit_dis: 0.460723, reader_cost: 4.61426 s, batch_cost: 6.85204 s, samples: 5120, ips: 74.72232 [2021/06/30 10:48:18] root INFO: epoch: [1/500], iter: 200, lr: 0.001000, loss: 29.476135, acc: 0.066406, norm_edit_dis: 0.463056, reader_cost: 7.58683 s, batch_cost: 9.89736 s, samples: 5120, ips: 51.73096 [2021/06/30 10:50:34] root INFO: epoch: [1/500], iter: 210, lr: 0.001000, loss: 29.196236, acc: 0.067383, norm_edit_dis: 0.462320, reader_cost: 8.02476 s, batch_cost: 10.27828 s, samples: 5120, ips: 49.81378 [2021/06/30 10:52:20] root INFO: epoch: [1/500], iter: 220, lr: 0.001000, loss: 28.938604, acc: 0.063477, norm_edit_dis: 0.458144, reader_cost: 5.67016 s, batch_cost: 7.69411 s, samples: 5120, ips: 66.54444 [2021/06/30 10:54:00] root INFO: epoch: [1/500], iter: 230, lr: 0.001000, loss: 29.047894, acc: 0.062500, norm_edit_dis: 0.455604, reader_cost: 5.10638 s, batch_cost: 7.11311 s, samples: 5120, ips: 71.97981 [2021/06/30 10:56:23] root INFO: epoch: [1/500], iter: 240, lr: 0.001000, loss: 29.001678, acc: 0.064453, norm_edit_dis: 0.455803, reader_cost: 9.38144 s, batch_cost: 11.40148 s, samples: 5120, ips: 44.90644 [2021/06/30 10:58:29] root INFO: epoch: [1/500], iter: 250, lr: 0.001000, loss: 29.021385, acc: 0.064453, norm_edit_dis: 0.457864, reader_cost: 7.79869 s, batch_cost: 9.81200 s, samples: 5120, ips: 52.18100 [2021/06/30 11:00:16] root INFO: epoch: [1/500], iter: 260, lr: 0.001000, loss: 29.368792, acc: 0.064453, norm_edit_dis: 0.456196, reader_cost: 5.77732 s, batch_cost: 7.78389 s, samples: 5120, ips: 65.77686 [2021/06/30 11:01:58] root INFO: epoch: [1/500], iter: 270, lr: 0.001000, loss: 29.761681, acc: 0.058594, norm_edit_dis: 0.449271, reader_cost: 5.33553 s, batch_cost: 7.35235 s, samples: 5120, ips: 69.63758 [2021/06/30 11:04:40] root INFO: epoch: [1/500], iter: 280, lr: 0.001000, loss: 28.928726, acc: 0.058594, norm_edit_dis: 0.451067, reader_cost: 11.32355 s, batch_cost: 13.34461 s, samples: 5120, ips: 38.36756 [2021/06/30 11:06:23] root INFO: epoch: [1/500], iter: 290, lr: 0.001000, loss: 28.899677, acc: 0.062500, norm_edit_dis: 0.457633, reader_cost: 5.42592 s, batch_cost: 7.42733 s, samples: 5120, ips: 68.93462 [2021/06/30 11:08:06] root INFO: epoch: [1/500], iter: 300, lr: 0.001000, loss: 28.954216, acc: 0.064453, norm_edit_dis: 0.461166, reader_cost: 5.36566 s, batch_cost: 7.37086 s, samples: 5120, ips: 69.46268 [2021/06/30 11:09:55] root INFO: epoch: [1/500], iter: 310, lr: 0.001000, loss: 28.804794, acc: 0.066406, norm_edit_dis: 0.459190, reader_cost: 6.03951 s, batch_cost: 8.04818 s, samples: 5120, ips: 63.61684 [2021/06/30 11:12:39] root INFO: epoch: [1/500], iter: 320, lr: 0.001000, loss: 28.737762, acc: 0.064453, norm_edit_dis: 0.455933, reader_cost: 11.53702 s, batch_cost: 13.57055 s, samples: 5120, ips: 37.72876 [2021/06/30 11:14:30] root INFO: epoch: [1/500], iter: 330, lr: 0.001000, loss: 28.733232, acc: 0.065430, norm_edit_dis: 0.462456, reader_cost: 6.19043 s, batch_cost: 8.20546 s, samples: 5120, ips: 62.39749 [2021/06/30 11:16:15] root INFO: epoch: [1/500], iter: 340, lr: 0.001000, loss: 28.733232, acc: 0.067383, norm_edit_dis: 0.463902, reader_cost: 5.59001 s, batch_cost: 7.60060 s, samples: 5120, ips: 67.36306 [2021/06/30 11:17:57] root INFO: epoch: [1/500], iter: 350, lr: 0.001000, loss: 28.968098, acc: 0.062500, norm_edit_dis: 0.449038, reader_cost: 5.36341 s, batch_cost: 7.37285 s, samples: 5120, ips: 69.44393 [2021/06/30 11:20:21] root INFO: epoch: [1/500], iter: 360, lr: 0.001000, loss: 29.146383, acc: 0.063477, norm_edit_dis: 0.450650, reader_cost: 9.52700 s, batch_cost: 11.55431 s, samples: 5120, ips: 44.31247 [2021/06/30 11:22:22] root INFO: epoch: [1/500], iter: 370, lr: 0.001000, loss: 29.051979, acc: 0.066406, norm_edit_dis: 0.450803, reader_cost: 7.20553 s, batch_cost: 9.21864 s, samples: 5120, ips: 55.53965 [2021/06/30 11:24:12] root INFO: epoch: [1/500], iter: 380, lr: 0.001000, loss: 28.745113, acc: 0.061523, norm_edit_dis: 0.450217, reader_cost: 6.10408 s, batch_cost: 8.12954 s, samples: 5120, ips: 62.98018 [2021/06/30 11:25:58] root INFO: epoch: [1/500], iter: 390, lr: 0.001000, loss: 29.158270, acc: 0.060547, norm_edit_dis: 0.456435, reader_cost: 5.68604 s, batch_cost: 7.71055 s, samples: 5120, ips: 66.40254 [2021/06/30 11:28:24] root INFO: epoch: [1/500], iter: 400, lr: 0.001000, loss: 29.241486, acc: 0.062500, norm_edit_dis: 0.459944, reader_cost: 9.71734 s, batch_cost: 11.73385 s, samples: 5120, ips: 43.63445 [2021/06/30 11:30:33] root INFO: epoch: [1/500], iter: 410, lr: 0.001000, loss: 29.073112, acc: 0.063477, norm_edit_dis: 0.461218, reader_cost: 8.07536 s, batch_cost: 10.09336 s, samples: 5120, ips: 50.72641 [2021/06/30 11:32:24] root INFO: epoch: [1/500], iter: 420, lr: 0.001000, loss: 29.064835, acc: 0.070312, norm_edit_dis: 0.461727, reader_cost: 6.17176 s, batch_cost: 8.18808 s, samples: 5120, ips: 62.52995 [2021/06/30 11:34:10] root INFO: epoch: [1/500], iter: 430, lr: 0.001000, loss: 29.064835, acc: 0.070312, norm_edit_dis: 0.464027, reader_cost: 5.75984 s, batch_cost: 7.77313 s, samples: 5120, ips: 65.86797 [2021/06/30 11:36:21] root INFO: epoch: [1/500], iter: 440, lr: 0.001000, loss: 28.942097, acc: 0.066406, norm_edit_dis: 0.460420, reader_cost: 8.17048 s, batch_cost: 10.17095 s, samples: 5120, ips: 50.33945 [2021/06/30 11:38:51] root INFO: epoch: [1/500], iter: 450, lr: 0.001000, loss: 28.776918, acc: 0.066406, norm_edit_dis: 0.459700, reader_cost: 10.11811 s, batch_cost: 12.11821 s, samples: 5120, ips: 42.25046 [2021/06/30 11:40:31] root INFO: epoch: [1/500], iter: 460, lr: 0.001000, loss: 28.640152, acc: 0.066406, norm_edit_dis: 0.464868, reader_cost: 5.17719 s, batch_cost: 7.19007 s, samples: 5120, ips: 71.20930 [2021/06/30 11:42:16] root INFO: epoch: [1/500], iter: 470, lr: 0.001000, loss: 28.173771, acc: 0.067383, norm_edit_dis: 0.464868, reader_cost: 5.62408 s, batch_cost: 7.63949 s, samples: 5120, ips: 67.02021 [2021/06/30 11:44:29] root INFO: epoch: [1/500], iter: 480, lr: 0.001000, loss: 28.759075, acc: 0.067383, norm_edit_dis: 0.464287, reader_cost: 8.42004 s, batch_cost: 10.42875 s, samples: 5120, ips: 49.09505 [2021/06/30 11:46:47] root INFO: epoch: [1/500], iter: 490, lr: 0.001000, loss: 28.696434, acc: 0.070312, norm_edit_dis: 0.466792, reader_cost: 8.84879 s, batch_cost: 10.86693 s, samples: 5120, ips: 47.11541 [2021/06/30 11:48:34] root INFO: epoch: [1/500], iter: 500, lr: 0.001000, loss: 28.362125, acc: 0.070312, norm_edit_dis: 0.472159, reader_cost: 5.78696 s, batch_cost: 7.79431 s, samples: 5120, ips: 65.68893 [2021/06/30 11:50:17] root INFO: epoch: [1/500], iter: 510, lr: 0.001000, loss: 28.226355, acc: 0.066406, norm_edit_dis: 0.473989, reader_cost: 5.42195 s, batch_cost: 7.42955 s, samples: 5120, ips: 68.91397 [2021/06/30 11:52:28] root INFO: epoch: [1/500], iter: 520, lr: 0.001000, loss: 28.292847, acc: 0.066406, norm_edit_dis: 0.470583, reader_cost: 8.22645 s, batch_cost: 10.24074 s, samples: 5120, ips: 49.99641 [2021/06/30 11:54:40] root INFO: epoch: [1/500], iter: 530, lr: 0.001000, loss: 28.736652, acc: 0.071289, norm_edit_dis: 0.468598, reader_cost: 8.24087 s, batch_cost: 10.27221 s, samples: 5120, ips: 49.84320 [2021/06/30 11:56:20] root INFO: epoch: [1/500], iter: 540, lr: 0.001000, loss: 28.425404, acc: 0.072266, norm_edit_dis: 0.473481, reader_cost: 5.15648 s, batch_cost: 7.16887 s, samples: 5120, ips: 71.41987 [2021/06/30 11:58:05] root INFO: epoch: [1/500], iter: 550, lr: 0.001000, loss: 28.159788, acc: 0.072266, norm_edit_dis: 0.471328, reader_cost: 5.54386 s, batch_cost: 7.55497 s, samples: 5120, ips: 67.76997 [2021/06/30 12:00:35] root INFO: epoch: [1/500], iter: 560, lr: 0.001000, loss: 28.237759, acc: 0.076172, norm_edit_dis: 0.471328, reader_cost: 10.07634 s, batch_cost: 12.09368 s, samples: 5120, ips: 42.33615 [2021/06/30 12:02:34] root INFO: epoch: [1/500], iter: 570, lr: 0.001000, loss: 28.269911, acc: 0.074219, norm_edit_dis: 0.467947, reader_cost: 7.06222 s, batch_cost: 9.07959 s, samples: 5120, ips: 56.39021 [2021/06/30 12:04:22] root INFO: epoch: [1/500], iter: 580, lr: 0.001000, loss: 28.190353, acc: 0.066406, norm_edit_dis: 0.467210, reader_cost: 5.86674 s, batch_cost: 7.88063 s, samples: 5120, ips: 64.96946 [2021/06/30 12:06:09] root INFO: epoch: [1/500], iter: 590, lr: 0.001000, loss: 27.891895, acc: 0.065430, norm_edit_dis: 0.470969, reader_cost: 5.81250 s, batch_cost: 7.82621 s, samples: 5120, ips: 65.42120 [2021/06/30 12:08:51] root INFO: epoch: [1/500], iter: 600, lr: 0.001000, loss: 27.933929, acc: 0.066406, norm_edit_dis: 0.477819, reader_cost: 11.27344 s, batch_cost: 13.28886 s, samples: 5120, ips: 38.52850 [2021/06/30 12:10:48] root INFO: epoch: [1/500], iter: 610, lr: 0.001000, loss: 28.078323, acc: 0.072266, norm_edit_dis: 0.474948, reader_cost: 6.84774 s, batch_cost: 8.84874 s, samples: 5120, ips: 57.86134 [2021/06/30 12:12:30] root INFO: epoch: [1/500], iter: 620, lr: 0.001000, loss: 28.703627, acc: 0.072266, norm_edit_dis: 0.466865, reader_cost: 5.39026 s, batch_cost: 7.39218 s, samples: 5120, ips: 69.26243 [2021/06/30 12:14:18] root INFO: epoch: [1/500], iter: 630, lr: 0.001000, loss: 28.899885, acc: 0.069336, norm_edit_dis: 0.459304, reader_cost: 5.94170 s, batch_cost: 7.95324 s, samples: 5120, ips: 64.37625 [2021/06/30 12:16:56] root INFO: epoch: [1/500], iter: 640, lr: 0.001000, loss: 28.561722, acc: 0.067383, norm_edit_dis: 0.461272, reader_cost: 10.87176 s, batch_cost: 12.88228 s, samples: 5120, ips: 39.74452 [2021/06/30 12:18:51] root INFO: epoch: [1/500], iter: 650, lr: 0.001000, loss: 28.615982, acc: 0.065430, norm_edit_dis: 0.461272, reader_cost: 6.67661 s, batch_cost: 8.68191 s, samples: 5120, ips: 58.97323 [2021/06/30 12:20:39] root INFO: epoch: [1/500], iter: 660, lr: 0.001000, loss: 28.805407, acc: 0.064453, norm_edit_dis: 0.460906, reader_cost: 5.85671 s, batch_cost: 7.87161 s, samples: 5120, ips: 65.04387 [2021/06/30 12:22:28] root INFO: epoch: [1/500], iter: 670, lr: 0.001000, loss: 28.465267, acc: 0.069336, norm_edit_dis: 0.470795, reader_cost: 6.01108 s, batch_cost: 8.02061 s, samples: 5120, ips: 63.83558 [2021/06/30 12:24:47] root INFO: epoch: [1/500], iter: 680, lr: 0.001000, loss: 28.128532, acc: 0.067383, norm_edit_dis: 0.468589, reader_cost: 8.96983 s, batch_cost: 10.99364 s, samples: 5120, ips: 46.57236 [2021/06/30 12:26:40] root INFO: epoch: [1/500], iter: 690, lr: 0.001000, loss: 28.108297, acc: 0.064453, norm_edit_dis: 0.461884, reader_cost: 6.44312 s, batch_cost: 8.46104 s, samples: 5120, ips: 60.51264 [2021/06/30 12:28:22] root INFO: epoch: [1/500], iter: 700, lr: 0.001000, loss: 28.473877, acc: 0.063477, norm_edit_dis: 0.463387, reader_cost: 5.25148 s, batch_cost: 7.26671 s, samples: 5120, ips: 70.45830 [2021/06/30 12:30:13] root INFO: epoch: [1/500], iter: 710, lr: 0.001000, loss: 28.715656, acc: 0.065430, norm_edit_dis: 0.464576, reader_cost: 6.30455 s, batch_cost: 8.31207 s, samples: 5120, ips: 61.59716 [2021/06/30 12:32:33] root INFO: epoch: [1/500], iter: 720, lr: 0.001000, loss: 28.714520, acc: 0.068359, norm_edit_dis: 0.468680, reader_cost: 9.09563 s, batch_cost: 11.10394 s, samples: 5120, ips: 46.10974 [2021/06/30 12:34:52] root INFO: epoch: [1/500], iter: 730, lr: 0.001000, loss: 28.349289, acc: 0.070312, norm_edit_dis: 0.468098, reader_cost: 9.05203 s, batch_cost: 11.05918 s, samples: 5120, ips: 46.29636 [2021/06/30 12:36:31] root INFO: epoch: [1/500], iter: 740, lr: 0.001000, loss: 28.157452, acc: 0.068359, norm_edit_dis: 0.466138, reader_cost: 5.03931 s, batch_cost: 7.04134 s, samples: 5120, ips: 72.71340 [2021/06/30 12:38:15] root INFO: epoch: [1/500], iter: 750, lr: 0.001000, loss: 28.477364, acc: 0.066406, norm_edit_dis: 0.466440, reader_cost: 5.44301 s, batch_cost: 7.45433 s, samples: 5120, ips: 68.68490 [2021/06/30 12:40:25] root INFO: epoch: [1/500], iter: 760, lr: 0.001000, loss: 28.604227, acc: 0.066406, norm_edit_dis: 0.468272, reader_cost: 8.13835 s, batch_cost: 10.15402 s, samples: 5120, ips: 50.42339 [2021/06/30 12:42:48] root INFO: epoch: [1/500], iter: 770, lr: 0.001000, loss: 28.356033, acc: 0.067383, norm_edit_dis: 0.469472, reader_cost: 9.39149 s, batch_cost: 11.40433 s, samples: 5120, ips: 44.89522 [2021/06/30 12:44:31] root INFO: epoch: [1/500], iter: 780, lr: 0.001000, loss: 28.179569, acc: 0.068359, norm_edit_dis: 0.468194, reader_cost: 5.44025 s, batch_cost: 7.44190 s, samples: 5120, ips: 68.79964 [2021/06/30 12:46:19] root INFO: epoch: [1/500], iter: 790, lr: 0.001000, loss: 28.387901, acc: 0.069336, norm_edit_dis: 0.468575, reader_cost: 5.89421 s, batch_cost: 7.91266 s, samples: 5120, ips: 64.70645 [2021/06/30 12:48:28] root INFO: epoch: [1/500], iter: 800, lr: 0.001000, loss: 28.387901, acc: 0.074219, norm_edit_dis: 0.470769, reader_cost: 7.97965 s, batch_cost: 9.99731 s, samples: 5120, ips: 51.21377 [2021/06/30 12:50:35] root INFO: epoch: [1/500], iter: 810, lr: 0.001000, loss: 27.823750, acc: 0.076172, norm_edit_dis: 0.471394, reader_cost: 7.84990 s, batch_cost: 9.85493 s, samples: 5120, ips: 51.95371 [2021/06/30 12:52:26] root INFO: epoch: [1/500], iter: 820, lr: 0.001000, loss: 28.115864, acc: 0.073242, norm_edit_dis: 0.470556, reader_cost: 6.19001 s, batch_cost: 8.21699 s, samples: 5120, ips: 62.30991 [2021/06/30 12:54:09] root INFO: epoch: [1/500], iter: 830, lr: 0.001000, loss: 28.710503, acc: 0.072266, norm_edit_dis: 0.469747, reader_cost: 5.46389 s, batch_cost: 7.47894 s, samples: 5120, ips: 68.45886 [2021/06/30 12:56:23] root INFO: epoch: [1/500], iter: 840, lr: 0.001000, loss: 29.178604, acc: 0.073242, norm_edit_dis: 0.467639, reader_cost: 8.52222 s, batch_cost: 10.52756 s, samples: 5120, ips: 48.63423 [2021/06/30 12:58:39] root INFO: epoch: [1/500], iter: 850, lr: 0.001000, loss: 28.458214, acc: 0.070312, norm_edit_dis: 0.470220, reader_cost: 8.67604 s, batch_cost: 10.69119 s, samples: 5120, ips: 47.88989 [2021/06/30 13:00:27] root INFO: epoch: [1/500], iter: 860, lr: 0.001000, loss: 27.963924, acc: 0.068359, norm_edit_dis: 0.470177, reader_cost: 5.88580 s, batch_cost: 7.91773 s, samples: 5120, ips: 64.66500 [2021/06/30 13:02:13] root INFO: epoch: [1/500], iter: 870, lr: 0.001000, loss: 29.029671, acc: 0.073242, norm_edit_dis: 0.466783, reader_cost: 5.70341 s, batch_cost: 7.71282 s, samples: 5120, ips: 66.38298 [2021/06/30 13:04:19] root INFO: epoch: [1/500], iter: 880, lr: 0.001000, loss: 28.463211, acc: 0.068359, norm_edit_dis: 0.464594, reader_cost: 7.62302 s, batch_cost: 9.63667 s, samples: 5120, ips: 53.13039 [2021/06/30 13:06:29] root INFO: epoch: [1/500], iter: 890, lr: 0.001000, loss: 28.166525, acc: 0.068359, norm_edit_dis: 0.466103, reader_cost: 8.12645 s, batch_cost: 10.13842 s, samples: 5120, ips: 50.50098 [2021/06/30 13:08:10] root INFO: epoch: [1/500], iter: 900, lr: 0.001000, loss: 28.250711, acc: 0.069336, norm_edit_dis: 0.465625, reader_cost: 5.14691 s, batch_cost: 7.16375 s, samples: 5120, ips: 71.47092 [2021/06/30 13:09:53] root INFO: epoch: [1/500], iter: 910, lr: 0.001000, loss: 28.285503, acc: 0.063477, norm_edit_dis: 0.465499, reader_cost: 5.39467 s, batch_cost: 7.40608 s, samples: 5120, ips: 69.13239 [2021/06/30 13:12:14] root INFO: epoch: [1/500], iter: 920, lr: 0.001000, loss: 28.497967, acc: 0.068359, norm_edit_dis: 0.468391, reader_cost: 9.28608 s, batch_cost: 11.30375 s, samples: 5120, ips: 45.29472 [2021/06/30 13:14:20] root INFO: epoch: [1/500], iter: 930, lr: 0.001000, loss: 28.180235, acc: 0.071289, norm_edit_dis: 0.470724, reader_cost: 7.59703 s, batch_cost: 9.63173 s, samples: 5120, ips: 53.15763 [2021/06/30 13:16:04] root INFO: epoch: [1/500], iter: 940, lr: 0.001000, loss: 27.983078, acc: 0.072266, norm_edit_dis: 0.466667, reader_cost: 5.58320 s, batch_cost: 7.60704 s, samples: 5120, ips: 67.30606 [2021/06/30 13:17:49] root INFO: epoch: [1/500], iter: 950, lr: 0.001000, loss: 27.746002, acc: 0.070312, norm_edit_dis: 0.466667, reader_cost: 5.60527 s, batch_cost: 7.61596 s, samples: 5120, ips: 67.22727 [2021/06/30 13:20:21] root INFO: epoch: [1/500], iter: 960, lr: 0.001000, loss: 28.008808, acc: 0.070312, norm_edit_dis: 0.471570, reader_cost: 10.29654 s, batch_cost: 12.31794 s, samples: 5120, ips: 41.56539 [2021/06/30 13:22:15] root INFO: epoch: [1/500], iter: 970, lr: 0.001000, loss: 28.126303, acc: 0.072266, norm_edit_dis: 0.473188, reader_cost: 6.43201 s, batch_cost: 8.45166 s, samples: 5120, ips: 60.57979 [2021/06/30 13:24:02] root INFO: epoch: [1/500], iter: 980, lr: 0.001000, loss: 28.711082, acc: 0.069336, norm_edit_dis: 0.468217, reader_cost: 5.77872 s, batch_cost: 7.79508 s, samples: 5120, ips: 65.68242 [2021/06/30 13:25:42] root INFO: epoch: [1/500], iter: 990, lr: 0.001000, loss: 28.707054, acc: 0.066406, norm_edit_dis: 0.460275, reader_cost: 5.10476 s, batch_cost: 7.12058 s, samples: 5120, ips: 71.90430 [2021/06/30 13:28:16] root INFO: epoch: [1/500], iter: 1000, lr: 0.001000, loss: 28.655977, acc: 0.063477, norm_edit_dis: 0.460463, reader_cost: 10.54009 s, batch_cost: 12.55472 s, samples: 5120, ips: 40.78149 [2021/06/30 13:29:54] root INFO: epoch: [1/500], iter: 1010, lr: 0.001000, loss: 28.081484, acc: 0.066406, norm_edit_dis: 0.463683, reader_cost: 4.90912 s, batch_cost: 6.93585 s, samples: 5120, ips: 73.81941 [2021/06/30 13:31:42] root INFO: epoch: [1/500], iter: 1020, lr: 0.001000, loss: 28.431808, acc: 0.066406, norm_edit_dis: 0.467774, reader_cost: 5.87391 s, batch_cost: 7.87105 s, samples: 5120, ips: 65.04853 [2021/06/30 13:33:25] root INFO: epoch: [1/500], iter: 1030, lr: 0.001000, loss: 28.702068, acc: 0.066406, norm_edit_dis: 0.466427, reader_cost: 5.50698 s, batch_cost: 7.51992 s, samples: 5120, ips: 68.08580 [2021/06/30 13:35:48] root INFO: epoch: [1/500], iter: 1040, lr: 0.001000, loss: 28.051796, acc: 0.068359, norm_edit_dis: 0.467071, reader_cost: 9.37153 s, batch_cost: 11.38322 s, samples: 5120, ips: 44.97849 [2021/06/30 13:38:00] root INFO: epoch: [1/500], iter: 1050, lr: 0.001000, loss: 27.695133, acc: 0.072266, norm_edit_dis: 0.469461, reader_cost: 8.25736 s, batch_cost: 10.26917 s, samples: 5120, ips: 49.85796 [2021/06/30 13:39:46] root INFO: epoch: [1/500], iter: 1060, lr: 0.001000, loss: 27.637615, acc: 0.072266, norm_edit_dis: 0.477125, reader_cost: 5.73387 s, batch_cost: 7.73790 s, samples: 5120, ips: 66.16784 [2021/06/30 13:41:32] root INFO: epoch: [1/500], iter: 1070, lr: 0.001000, loss: 28.487730, acc: 0.070312, norm_edit_dis: 0.474837, reader_cost: 5.65326 s, batch_cost: 7.67167 s, samples: 5120, ips: 66.73904 [2021/06/30 13:43:31] root INFO: epoch: [1/500], iter: 1080, lr: 0.001000, loss: 28.685019, acc: 0.069336, norm_edit_dis: 0.472277, reader_cost: 7.01997 s, batch_cost: 9.02523 s, samples: 5120, ips: 56.72987 [2021/06/30 13:46:16] root INFO: epoch: [1/500], iter: 1090, lr: 0.001000, loss: 28.638580, acc: 0.072266, norm_edit_dis: 0.475241, reader_cost: 11.71552 s, batch_cost: 13.70732 s, samples: 5120, ips: 37.35229 [2021/06/30 13:48:01] root INFO: epoch: [1/500], iter: 1100, lr: 0.001000, loss: 28.413712, acc: 0.069336, norm_edit_dis: 0.470005, reader_cost: 5.62880 s, batch_cost: 7.64920 s, samples: 5120, ips: 66.93507 [2021/06/30 13:49:38] root INFO: epoch: [1/500], iter: 1110, lr: 0.001000, loss: 28.413712, acc: 0.070312, norm_edit_dis: 0.466420, reader_cost: 4.80654 s, batch_cost: 6.81625 s, samples: 5120, ips: 75.11465 [2021/06/30 13:51:40] root INFO: epoch: [1/500], iter: 1120, lr: 0.001000, loss: 28.347948, acc: 0.069336, norm_edit_dis: 0.472470, reader_cost: 7.23361 s, batch_cost: 9.25127 s, samples: 5120, ips: 55.34376 [2021/06/30 13:54:09] root INFO: epoch: [1/500], iter: 1130, lr: 0.001000, loss: 28.396919, acc: 0.069336, norm_edit_dis: 0.468906, reader_cost: 10.00217 s, batch_cost: 12.03964 s, samples: 5120, ips: 42.52620 [2021/06/30 13:55:53] root INFO: epoch: [1/500], iter: 1140, lr: 0.001000, loss: 29.213041, acc: 0.070312, norm_edit_dis: 0.453394, reader_cost: 5.57170 s, batch_cost: 7.57531 s, samples: 5120, ips: 67.58799 [2021/06/30 13:57:43] root INFO: epoch: [1/500], iter: 1150, lr: 0.001000, loss: 28.723248, acc: 0.065430, norm_edit_dis: 0.458687, reader_cost: 6.08701 s, batch_cost: 8.09834 s, samples: 5120, ips: 63.22282 [2021/06/30 13:59:42] root INFO: epoch: [1/500], iter: 1160, lr: 0.001000, loss: 28.491301, acc: 0.066406, norm_edit_dis: 0.468724, reader_cost: 7.07698 s, batch_cost: 9.08576 s, samples: 5120, ips: 56.35192 [

大佬，第二次看见你了，您的数据集是怎么得到的呢，以及所用的字典是自己写的还是官方的呢