tks for your information, I don't know the cause of this error, so I added a little note for this issue in the readme

[luckystar1992]

tks for your information, I don't know the cause of this error, so I added a little note for this issue in the readme

Originally posted by @WeiNyn in #5 (comment)

[luckystar1992]

Facing the same issue with > 4.2, now the issue is File "***/lib/python3.6/site-packages/transformers/trainer_pt_utils.py", line 48, in torch_pad_and_concatenate
if len(tensor1.shape) == 1 or tensor1.shape[1] == tensor2.shape[1]:
IndexError: tuple index out of range

In the forward() function of classifier.py, it returns a dict, which loss is a Tensor with shape 0. It leads the error above, so is there any good ways for that.

[luckystar1992]

This error only appears in my mac mini m1, and I have changed to a Linux platform and it works, but have to set output_attentions and output_hidden_states to True in the classifier.py, it avoids two NoneType items in the final return