用的默认rec_chinese_lite_v2.0
输出就没出现过ACC
[2022/04/09 17:36:17] root INFO: train with paddle 2.2.2 and device CUDAPlace(0)
[2022/04/09 17:36:17] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:17] root INFO: Initialize indexs of datasets:['./train_data/val_list.txt']
[2022/04/09 17:36:19] root INFO: load pretrain successful from ./pretrain_models/best_accuracy
[2022/04/09 17:36:19] root INFO: train dataloader has 1 iters
[2022/04/09 17:36:19] root INFO: valid dataloader has 1 iters
[2022/04/09 17:36:19] root INFO: During the training process, after the 0th iteration, an evaluation is run every 2000 iterations
[2022/04/09 17:36:19] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:19] root INFO: save model in ./output/rec_chinese_lite_v2.0\latest
[2022/04/09 17:36:19] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:19] root INFO: save model in ./output/rec_chinese_lite_v2.0\latest
[2022/04/09 17:36:19] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:19] root INFO: save model in ./output/rec_chinese_lite_v2.0\latest
[2022/04/09 17:36:19] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:20] root INFO: save model in ./output/rec_chinese_lite_v2.0\latest
[2022/04/09 17:36:20] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:20] root INFO: save model in ./output/rec_chinese_lite_v2.0\latest
[2022/04/09 17:36:20] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:20] root INFO: save model in ./output/rec_chinese_lite_v2.0\latest
[2022/04/09 17:36:20] root INFO: Initialize indexs of datasets:['./train_data/train_list.txt']
[2022/04/09 17:36:20] root INFO: save model in ./output/rec_chinese_lite_v2.0\latest
是不是数据集处理的问题呢，可以先替换原数据集的部分数据试试
好 我试试 另外测试数据集的图片是不是要统一大小？ 还是无所谓
有结果了吗？是数据问题吗？我也出现了这样的问题
在吗？有解决方案了吗？萌新求解答
训练多少代没出acc哪，如果数据量多的话，或者学习率小，十几代acc才开始变化也是正常的
你这个显示的 ，跟我不太一样呀
你评估一下
[2022/04/22 21:56:21] root INFO: metric eval ***************
[2022/04/22 21:56:21] root INFO: precision:0.7950859950859951
[2022/04/22 21:56:21] root INFO: recall:0.8983897834536368
[2022/04/22 21:56:21] root INFO: hmean:0.8435870698644421
[2022/04/22 21:56:21] root INFO: fps:10.895837072274382