edgarriba
Example Run from README leads to infinite loss error after a few iterations
Having checked my GPU setup and gotten the code to download the dataset and run, I get this
[0/100][0/733] Loss_D: 1.38686203956604 Loss_G: 1.3840898275375366
[0/100][1/733] Loss_D: 1.385105848312378 Loss_G: 1.3885846138000488
[0/100][2/733] Loss_D: 1.3828593492507935 Loss_G: 1.387503981590271
[0/100][3/733] Loss_D: 1.3838441371917725 Loss_G: 1.3841853141784668
[0/100][4/733] Loss_D: 1.6616134643554688 Loss_G: 1.4784884452819824
[0/100][5/733] Loss_D: 18.7115478515625 Loss_G: 14.966793060302734
Traceback (most recent call last):
File "main.py", line 328, in <module>
# call train/test routines
File "main.py", line 254, in train
D_loss = compute_loss(batch_size, d_loss=True)
File "main.py", line 170, in compute_loss
loss = torch.mean(softplus(-data_preds) + softplus(sample_preds))
RuntimeError: value cannot be converted to type double without overflow: inf
This is running the example cli command given in the README
