AttributeError: 'list' object has no attribute 'long' when forward with argument prev_output_tokens #1
Description
When running the pretrain script, in wav2seq/models/wav2vec2_asr_v2.py, Line 794
The docs in forward function mentioned that the prev_output_tokens is LongTensor with shape (batch, tgt_len), while we get with list of tensors with different tgt_len for each tensor.
Do anyone know how to solve it? I'm not sure whether setting the batch size to 1 and take the first element inside the list would work or not?