search

lucidrains / audiolm-pytorch

Implementation of AudioLM, a SOTA Language Modeling Approach to Audio Generation out of Google Research, in Pytorch
MIT License
2.32k stars 249 forks source link

About get_embeds function #269

Open jihoojung0106 opened 5 months ago

jihoojung0106 commented 5 months ago

in get_embeds function, default pad_id is set to -1. but, in CoarseTransformerWrapper, I remember that the padding was set to 0. in that case, shouldn't the get_embeds function in the CoarseTransformer foward function gets the argument pad_id=0??

lucidrains commented 5 months ago

@jihoojung0106 where do you see it being set to 0? i see -1 here https://github.com/lucidrains/audiolm-pytorch/blob/main/audiolm_pytorch/audiolm_pytorch.py#L1530