Open Duanener opened 8 months ago
Hello @Duanener, that's a good question. For the most part we used a code base very similar to that of the VQ-GAN codebase with some modifications from VQ-VAE-2, namely, the projection into a lower dimensional space for matching to the codebook elements.
Enhancement
Thanks for this wonderful work. However, is there any guidance on training the VQ-VAE in MS-ILLM ?