Error loading baseline weights

[FelipeMarra]

Loading baseline with its weights throws the following error:

RuntimeError: Error(s) in loading state_dict for CNN:
size mismatch for dense.weight: copying a param with shape torch.Size([56, 64]) from checkpoint, the shape in current model is torch.Size([15, 64]).
size mismatch for dense.bias: copying a param with shape torch.Size([56]) from checkpoint, the shape in current model is torch.Size([15]).

[FelipeMarra]

That actually can be fixed just by changing the number of classes to 56, indicating the baseline was trained only in the mood/theme subset.

Just one more thing, I created this fork, and it works fine in the training loop, but on validation I got some errors. The first one, I solved by modifying line 130 on the solver, because the data loader returns 3 values, not 2. The second one, which I'm still not able to fix, was:

ValueError: Only one class present in y_true. ROC AUC score is not defined in that case.

thrown by line 172 on the solver. Is this because the solver was only meant to work with the mood/theme subset, and for that subset every label contains more than 2 tags?

[dbogdanov]

The baseline has been provided by @minzwon, perhaps he can reply.