Vidacios
Results
2
comments of
Vidacios
使用non-dist多卡训练好像就有这个问题，换成dist多卡训练就好了 但不太清楚non-dist和dist这两种多卡训练有什么区别。。。感觉non-dist多卡像是把总的iters均分给每块gpu
I find that SwinIR regards as stage1 model in training, but BSRNet regards as stage1 model in inference. Why?