ser 模型微调是精度是0

[lxw316830]

[2024-03-29 17:28:43,426] [ INFO] - Already cached /root/.paddlenlp/models/layoutxlm-base-uncased/sentencepiece.bpe.model
[2024-03-29 17:28:43,819] [ INFO] - tokenizer config file saved in /root/.paddlenlp/models/layoutxlm-base-uncased/tokenizer_config.json
[2024-03-29 17:28:43,819] [ INFO] - Special tokens file saved in /root/.paddlenlp/models/layoutxlm-base-uncased/special_tokens_map.json
[2024/03/29 17:28:43] ppocr INFO: Initialize indexs of datasets:['/root/autodl-tmp/PaddleOCR/train_data/sfz/val.json']
[2024-03-29 17:28:43,820] [ INFO] - Already cached /root/.paddlenlp/models/layoutxlm-base-uncased/sentencepiece.bpe.model
[2024-03-29 17:28:44,204] [ INFO] - tokenizer config file saved in /root/.paddlenlp/models/layoutxlm-base-uncased/tokenizer_config.json
[2024-03-29 17:28:44,205] [ INFO] - Special tokens file saved in /root/.paddlenlp/models/layoutxlm-base-uncased/special_tokens_map.json
[2024-03-29 17:28:44,206] [ INFO] - Already cached /root/.paddlenlp/models/vi-layoutxlm-base-uncased/model_state.pdparams
W0329 17:28:44.210135 6339 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 8.9, Driver API Version: 12.2, Runtime API Version: 11.8
W0329 17:28:44.230378 6339 gpu_resources.cc:149] device: 0, cuDNN Version: 8.6.
[2024/03/29 17:28:45] ppocr INFO: train dataloader has 2 iters
[2024/03/29 17:28:45] ppocr INFO: valid dataloader has 1 iters
[2024/03/29 17:28:45] ppocr INFO: During the training process, after the 0th iteration, an evaluation is run every 19 iterations
[2024/03/29 17:28:47] ppocr INFO: epoch: [1/50], global_step: 2, lr: 0.000006, loss: 1.544584, avg_reader_cost: 0.18231 s, avg_batch_cost: 0.22472 s, avg_samples: 1.2, ips: 5.33998 samples/s, eta: 0:01:50
[2024/03/29 17:28:54] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:28:55] ppocr INFO: epoch: [2/50], global_step: 4, lr: 0.000019, loss: 1.298807, avg_reader_cost: 0.77165 s, avg_batch_cost: 0.78210 s, avg_samples: 1.2, ips: 1.53433 samples/s, eta: 0:04:01
[2024/03/29 17:29:03] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:29:05] ppocr INFO: epoch: [3/50], global_step: 6, lr: 0.000031, loss: 0.776025, avg_reader_cost: 0.91636 s, avg_batch_cost: 0.92626 s, avg_samples: 1.2, ips: 1.29554 samples/s, eta: 0:05:02
[2024/03/29 17:29:12] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:29:14] ppocr INFO: epoch: [4/50], global_step: 8, lr: 0.000043, loss: 0.333102, avg_reader_cost: 0.87588 s, avg_batch_cost: 0.88579 s, avg_samples: 1.2, ips: 1.35472 samples/s, eta: 0:05:24
[2024/03/29 17:29:22] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:29:23] ppocr INFO: epoch: [5/50], global_step: 10, lr: 0.000048, loss: 0.108488, avg_reader_cost: 0.91669 s, avg_batch_cost: 0.92675 s, avg_samples: 1.2, ips: 1.29484 samples/s, eta: 0:05:37
[2024/03/29 17:29:31] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:29:32] ppocr INFO: epoch: [6/50], global_step: 12, lr: 0.000048, loss: 0.019313, avg_reader_cost: 0.87929 s, avg_batch_cost: 0.88936 s, avg_samples: 1.2, ips: 1.34928 samples/s, eta: 0:05:39
[2024/03/29 17:29:39] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:29:41] ppocr INFO: epoch: [7/50], global_step: 14, lr: 0.000048, loss: 0.010023, avg_reader_cost: 0.86764 s, avg_batch_cost: 0.87723 s, avg_samples: 1.2, ips: 1.36795 samples/s, eta: 0:05:38
[2024/03/29 17:29:48] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:29:50] ppocr INFO: epoch: [8/50], global_step: 16, lr: 0.000047, loss: 0.005739, avg_reader_cost: 0.87828 s, avg_batch_cost: 0.88822 s, avg_samples: 1.2, ips: 1.35101 samples/s, eta: 0:05:36
[2024/03/29 17:29:57] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
[2024/03/29 17:29:59] ppocr INFO: epoch: [9/50], global_step: 18, lr: 0.000046, loss: 0.003596, avg_reader_cost: 0.85873 s, avg_batch_cost: 0.86907 s, avg_samples: 1.2, ips: 1.38079 samples/s, eta: 0:05:31
[2024/03/29 17:30:06] ppocr INFO: save model in ./output/ser_vi_layoutxlm_xfund_zh/latest
eval model:: 100%|█████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 1.68it/s]
[2024/03/29 17:30:08] ppocr INFO: cur metric, precision: 0.0, recall: 0.0, hmean: 0.0, fps: 87.63388933384407

精度一直是0

[zhangyubo0722]

从log看你的训练iter较小，尝试减小bs试试

[lxw316830]

那是0也不正常吧

[lxw316830]

有没有人帮忙看看，有偿的（17610296011）

[UserWangZz]

是不是数据量太小了，这类任务很吃数据的