LiPan123456789

When I execute the first command in python tools/valid.py \ --cfg experiments/coco/rescore_coco.yaml \ TEST.MODEL_FILE model/pose_coco/pose_dekr_hrnetw32.pth, I can't find '../data/rescore_data/rescore_dataset_train_coco_kpt' in it, and an error will be reported after the execution of more than 110000 pictures. I think it's because '../data/rescore_data/rescore_dataset_train_coco_kpt' needs to be created by myself, so I created such a file and modified the path in 'score/data/rescore_data/recore_dataset_train_coco_kpt', but an error is still reported. I hope you can help me solve it. In addition, I have another question. I have changed it three times. Each time I need to run 110000 pictures first, so it takes a lot of time. Is there any way to make him not need to run these 110000 pictures again?

2

command：  in rescore_coco.yaml： The path before modification is in the box, and the path after modification is in the ellipse：  I created this myself: 

When I was running the training of crowdpose dataset, there were 'nan' and 'NaN or Inf found in input tensor' problems. I checked the data. This problem is caused by the disappearance or explosion of the gradient. After I adjusted the learning rate, this problem still can not be solved. Please help me.

1

Although it has this problem, it is still running normally. Is it right to continue running like this? 

关于EPL作为开源的、统一了多种并行策略、灵活易用的自研分布式深度学习训练框架。

1

关于EPL的易用性、用户的模型编程接口和训练接口均基于TensorFlow。基于这点考虑，请问后续会出pytorch版本的吗，或者说tensorflow在这方面有什么优势，谢谢！