Repository Issue Activity (beta)

#225 Multi-Scale Residual Vector Quantization

#223 Rotation trick diverges when stochastic_sample_codes=True

#239 New SOTA Method (FVQ)

#241 Citing?

#240 FSQ returns OOB indices when noise_dropout > 0

#237 Very large shift for levels = 2

#238 commit_loss increases during training

#236 Learning codebooks also updates the encoder ?

#229 RuntimeError: expected scalar type Float but found BFloat16

#233 Dimension mismatch in Residual VQ

#232 balance the items in the leaf node of RQ

#230 Question Regarding Residual Subtraction in ResidualFSQ

#228 [bug] when dim=1

#227 Codebook diversity loss incorrect

#216 Which technique works?

#224 VectorQuantize.get_codes_from_indices() bug

#221 How to update EMA only at desired steps

#222 Learnable codebook synchronization across workers ?

#220 Is there a miscalculation in EuclideanCodebook's embed_avg?

#218 Channel-Split Quantization

#167 codebook update

#217 RVQ not converge

#106 nan encountered in `cdist()`

#110 How to train VectorQuantize codebook?

#26 EMA update on CosineCodebook