Koki Hokao
Koki Hokao
Hello @pacman100, is there any update on this? It seems that `save_state()` is called multiple times also when running with `distributed_type: MULTI_GPU`. When `automatic_checkpoint_naming=True`, I got the following same error....
@pacman100 I added `wait_for_everyone()` before `save_state()`, but it still raises the same error.
I've tried the experiment several times, but errors rarely occur, and in most cases, there seem to be no errors. It might be an issue with my server, so you...