Sounds good! Thanks for your help!
I can’t quite figure out why predictions are running slow. They seem to be running on CPU and consuming all the available RAM. Any ideas? Thanks!
Here is the code and associated outputs:
learn.dls.device
Out[50]:
device(type=‘cuda’, index=0)
In [51]:
next(learn.model.parameters()).is_cuda
Out[51]:
True
In [*]:
preds,targs = learn.get_preds(dl=tst.tls)
preds_clipped = preds.clamp(.0001, .999)
0.00% [0/1 00:00<00:00]