Closed jxust01 closed 2 years ago
Hi, please change device = 'cuda:2'
to device = 'cuda'
, and it should work fine.
Hi, please change
device = 'cuda:2'
todevice = 'cuda'
, and it should work fine.
thanks,it works,but why can't set the device = 'cuda:2'
?
Hi, please change
device = 'cuda:2'
todevice = 'cuda'
, and it should work fine.thanks,it works,but why can't set the
device = 'cuda:2'
?
I think it is a bug in the PyTorch custom CUDA kernel implementation. You can also use device='cuda2'
by adding FORCE_NATIVE=1
to the environmental variables (see here https://github.com/mit-han-lab/anycost-gan/blob/1177beafaf5ed486156f78f411c34e62f0d46f14/models/ops.py#L8)
Thank you very much for your answers, looking forward to creating awesome things with anycost-gan