codebook loss is 0 - Githubissues

duchenzhuang / FSQ-pytorch

A Pytorch Implementation of Finite Scalar Quantization

67 stars 4 forks source link

Open zyh-uaiaaaa opened 8 months ago

zyh-uaiaaaa commented 8 months ago

Hi,

Nice work! But I train the VQ code on my own dataset, the codebook loss is 0.0, is this correct?

duchenzhuang commented 5 months ago

In FSQ, there is no codebook loss, so 0.0 is right