about RuntimeError

[isihkr]

Hi!
In test phase, i got RuntimeError: CUDA out of memory. I have only a11G 2080ti GPU.
Could u please help me fix it?
@dypark86 @xairc

[Kwongrf]

I got the same error, I just reduce the sidelen from 144 to 112 and then retrain the model.

[isihkr]

Thank you! I'll try for it.

[isihkr]

hello！sorry to bother you again.
I didn't reduce the sidelen and used torch.cuda.empty_cache() to fix the error, like this:
try:
output = net(inputdata, inputcoord)
except RuntimeError as e:
if 'out of memory' in str(e):
print('|WARNING: ran out of memory')
if hasattr(torch.cuda, 'empty_cache'):
torch.cuda.empty_cache()
else:
raise e
in both train and test phase.
However, when i used the pretrained model and ran xai-viewer.py, i can't get the same predicted results as @dypark86 does.
Do you kown why?
@Kwongrf

[Kwongrf]

I can't get the same results after I reduced sidelen. Then I retrained the model with reduced sidelen ,I got similar results compared to the author. Sorry, I haven't tried your method yet.