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
44.38k stars 7.83k forks source link

为什么训练自己数据集,不会保存训练模型? #7545

Closed bank010 closed 2 years ago

bank010 commented 2 years ago

请提供下述完整信息以便快速定位问题/Please provide the following information to quickly locate the problem

andyjiang1116 commented 2 years ago

你运行指令是推理命令,不是训练指令

bank010 commented 2 years ago

GPU训练 支持单卡,多卡训练

训练日志会自动保存到 配置文件中"{Global.save_model_dir}" 下的train.log文件中

SER单卡训练

python3 tools/train.py -c configs/kie/vi_layoutxlm/ser_vi_layoutxlm_xfund_zh.yml

SER多卡训练,通过--gpus参数指定卡号

python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/kie/vi_layoutxlm/ser_vi_layoutxlm_xfund_zh.yml

RE任务单卡训练

python3 tools/train.py -c configs/kie/vi_layoutxlm/re_vi_layoutxlm_xfund_zh.yml

这些是训练指令吗?

andyjiang1116 commented 2 years ago

是的

bank010 commented 2 years ago

谢谢!