benatorc / otgnn Goto Github PK

The title says it all

Initialized model with 13572 params
0% 0/167 [00:00<?, ?it/s]/content/OTGNN/models/ot_modules.py💯 UserWarning: Implicit dimension choice for log_softmax has been deprecated. Change the call to include dim=X as an argument.
nce_reg = torch.nn.LogSoftmax()(torch.stack(all_nce_dists))[0]

and the code was
.... nce_reg = torch.nn.LogSoftmax()(torch.stack(all_nce_dists))[0]

So I manually changed it as
.... nce_reg = torch.nn.LogSoftmax(dim=0)(torch.stack(all_nce_dists))[0]
Not to see the error message and it actually worked but I know that this wouldn't work well in some condition.
Is there any reason to choice dimension between -1 or 0? and what should I select? Anyway I love your contents!

I found model_load util in OTGNN/utils/model_utils.py and loaded my model properly. And Now don't get how to stop my training and restart from checkpoint by loading pretrained model.