When I try to train with the new dev pytorch, it stops without giving an error message. The old version, i.e. 129, said something about gloo but it seems to have been fixed in the latest release, but it can’t train.
You would need to provide more information about your use case and the issue, ideally with a minimal and executable code snippet reproducing your issue.
I use a program called RVC voice. I have upgraded my python. When I use pytorch torch-2.7.1+cu128 I can train. I use a 4080 when I use torch-2.11.0.dev20260212+cu130 it stops after some time. Version 129 came up with a gloo error. and what is this meen?
C:\RVC20240604Nvidia50x0\infer\modules\train\train.py:486: FutureWarning: torch.cuda.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cuda', args...) instead.
with autocast(enabled=hps.train.fp16_run):
C:\RVC20240604Nvidia50x0\infer\modules\train\train.py:489: FutureWarning: torch.cuda.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cuda', args...) instead.
with autocast(enabled=False):