search

lucidrains / audiolm-pytorch

Implementation of AudioLM, a SOTA Language Modeling Approach to Audio Generation out of Google Research, in Pytorch
MIT License
2.33k stars 249 forks source link

Confusion about the coarse transformer trainer #167

Closed xtluo closed 1 year ago

xtluo commented 1 year ago

https://github.com/lucidrains/audiolm-pytorch/blob/241bae318f3e49860e765058c0c7abeceb1c9760/audiolm_pytorch/audiolm_pytorch.py#L1414

@lucidrains Here at line 1414 and 1415, why not:

batch = raw_wave_for_codec.shape[0]
num_timesteps = raw_wave_for_codec.shape[1]
lucidrains commented 1 year ago

@xtluo Thanks Xiaotao! https://github.com/lucidrains/audiolm-pytorch/commit/914ba79adbde1af937b8bb8e29e17e7b2e295dc2