IndexError: too many indices for array

[dineshiitkgp]

Hi,

I got the following error while training. (python train.py --ca) i guess in utils.py line 145 loss.data.cpu().numpy() is an empty array

can you please let us know how to resolve this issue.

[littlerookie]

can you tell me how did you solve the error?

[saxonelf]

just use pytorch with version under 0.4

[tomfat]

In PyTorch >=0.4. The return value of cpu() has been changed to 0-dimensional. So just modify the following line will fix this issue.

cum_loss += loss.data.cpu().numpy()[0]*(ed - st)

Change to

cum_loss += loss.data.cpu().numpy()*(ed - st)

[pprassu]

thanks tom. your solution is working :-)