vector-quantize-pytorch

vector-quantize-pytorch copied to clipboard

Why do I get almost the same codes after the 1st batch?

1

Hi there, I am trying to quantize my input feature `sparse_feat` with the following codes in my network. ``` class MyModel(nn.Module): def __init__ (self): super(MyModel, self).__init__() self.residual_vq = ResidualVQ( dim...

tanyz0208

[New quantization method] Binary Spherical Quantization

12

Binary Spherical Quantization is an extension of binarized Lookup-Free Quantization (from MagViT2), but they map the values not to a binarized hypercube, but on a sphere https://arxiv.org/abs/2406.07548 The authors claim...

kabachuha

The new version of FSQ, with the possibility to not compute indices and return only None, has a small bug: ```python from vector_quantize_pytorch.finite_scalar_quantization import FSQ import torch quantizer = FSQ(levels=[8,5,5,5],...

MisterBourbaki

Some clarification required

3

Hi Phil, I was experimenting with FSQ/LFQ for 3d-motion autoencoder and was wondering how to work through LFQ variety of options? With FSQ its quite straight forward as it does...

inspirit

HI, I am a rookie in this area, and I'm learning the original VQ-VAE. According to the original paper, how can I generate images using PixelCNN after finishing training the...

neilwen987

If I use a Grouped-VQ with two groups and codebook size =100, then the count of final speech tokens is 100 * 100 = 10000; If I use a 2-codebook...

JohnHerry

Difference between FSQ and LFQ

2

Hi, I was wondering about the difference between FSQ (finite scalar quantization) and LFQ (look-up free quantization) for vector quantization. As for a single vector among the encoder output, whatever...

JasonShen-SH

LatentQuantize exploding loss

2

I'm trying to use the LatentQuantize model in an autoencoder context. My inputs are flat 1-d tensors (32) and my encoder passes a shape of (batch_size, 64) to the quantizer....

jbmaxwell

Hi, I am interested in learning codewords (not using EMA) that are L2-normalized and orthonormal with each other. To do so, I created the vector quantizer using the following configuration:...

anupsingh15

begin to test out a population based vector quantizer with crossover from genetic algorit…

1

…hms. share the idea https://ieeexplore.ieee.org/document/764879 https://web3.arxiv.org/abs/2503.19037 https://github.com/lucidrains/genetic-algorithm-pytorch/blob/main/kmeans.py

lucidrains