Where is the weight of vqmodel's decoder?
#2
by
kowaides
- opened
How can I evaluate the reconstruction performance of a VQ model?
@kowaides the VQ decoder was not released by Meta team, and the current implementation supports only text generation. However, we have a different lab that trained a decoder module. There is a PR to track (https://github.com/huggingface/transformers/pull/32013), should be merged to Transformers soon