vq-vae-2-pytorch

vq-vae-2-pytorch copied to clipboard

how to do with video?

1

if i want to generate video, how to do with it?

1306107097

class-conditional samples

1

Thanks for implementing this in pytorch. At the current stage, is it possible to generate class-conditional samples? (Like in the ImageNet experiments of the paper)

edgarschnfld

Hi @rosinality, thank you a lot for your contribution. I have a question regarding the upsampling of the top latent representation. We use top decoder to upsample the quantized top...

space1panda

I am a little confused about the encoder and it seems that vectors are not compressed。 The question I want to ask is 1. which vector is latent space vector...

27yw

Similar to issue #43, I trained the VQ-VAE for about 37 epochs with a dataset of about 30k images. But the result is very blurry.  Do I need more...

easonoob

Hi, I've trained the VQVAE and the Pixel snail models on a dataset with multiple classes. However when I sample, I always seem to be getting different images of the...

parthnatekar

correctly initialize the moving counter of codebook entry utilization. initializing this to 0 leads to divides by ~0 and numerically unstable embeddings at the beginning of training.

provos

I want to use the pre trained weights first to encode that image and then decode the latent representation of that image?

asrays

why the bottom and top level quantizer have the same table size in the constructor of the main module

daleigehhh