Silky Singh
Silky Singh
@languandong You can use both, doesn't matter as long as `optimizer.zero_grad()` is called before `loss.backward()`. Note that `optimizer.zero_grad()` zeroes out the gradients in the `grad` field of the tensors, and...
https://stackoverflow.com/questions/8220702/error-int-object-is-not-subscriptable-python See if an integer variable has been subscripted in the code anywhere, like `a[idx]`, when actually `a` is just an integer
The `contain_dontcare_label` tells the model to discard a particular label in the semantic map. Basically, these are values in the semantic map whose correct label is not available in the...
Hi, I see this issue with the wav2vec2-base model. From the error log, it seems that the expected batch size is different. In the screenshot attached, could you try changing...
hi all, I got base model running with the following change in `wav2vec2_models.py`. Upon some googling, I found that base models are more rigid with the batch dimensions, hence causing...
closing this as the fix has been added in #2572
Do we want to add Audio Question Answering tasks too? (AQA is a task in which a system examines both an audio signal and a natural language question to produce...
fix has been added in PR #2572