I tried to train CE2P on Pytorch 0.5. But this error occurs:
dimension specified as 0 but tensor has no dimensions
To solve this, I modified forward() in CriterionDataParallel and added a dimension for each entry in outputs. Then the training works and the loss has successfully converged.
I opened this issue because the loss failed to converge on Pytorch 0.5. Now the training seems working fine. So I think this issue should be closed. Thanks a lot!
I tried to train CE2P on Pytorch 0.5. But this error occurs:
To solve this, I modified
forward()
inCriterionDataParallel
and added a dimension for each entry inoutputs
. Then the training works and the loss has successfully converged.I opened this issue because the loss failed to converge on Pytorch 0.5. Now the training seems working fine. So I think this issue should be closed. Thanks a lot!