Question aabout KFoldDataset and kfold-cross-valid.py

[BingliangLi]

Hi, I would like to use 5 fold cross-validation to train my model, I found there are two possible implantations:

• Use KFoldDataset, and manually repeat 5 times
• Use kfold-cross-valid.py, only need to run it once

But I'm not sure if kfold-cross-valid.py support multi-GPU training, I can't find any related documentation, and by the script's content, I think it probably does not. I want to ask if I want to perform a 5-fold cross-validation with multiple GPUs, is using KFoldDataset the only way?

[mzr1996]

Hello, which version are you using? 1.x or master?
I will try to test the multi-gpu training for k-fold validation recently

[BingliangLi]

Hi, I'm using the lastest version.

[mzr1996]

Ok, welcome to use our 1.x branch, which will become the master branch at the end of this year. And of course I will also test it on the master branch.