meshgpt-pytorch

meshgpt-pytorch copied to clipboard

[Critical] Very high loss rate at first few tokens (classifier free guidance not working)

34

@lucidrains This is a issue I'm having a while, the cross-attention is very weak at the start of the sequence. When the transformer starts with no tokens it will relay...

MarcusLoppe

Transformer keeps predicting the same token

6

Hi @lucidrains and @MarcusLoppe, I have succefully trained a meshAutoencoder, where validation loss is as low as training loss. I used around 28000 meshes from shapenet for training and validation....

MagicianWu

Added train by epoch for Trainer and added support for texts

7

Trainer - Added train function that will train with epochs instead of steps - Added option to display graph (maybe remove since it requires matlib?) - Added checkpoint_every_epoch to init...

MarcusLoppe

Proper dataset that reduces VRAM usage and provides higher performance.

14

I've created a dataset class which hopefully help beginners. Features: * Save & load compressed data in .npz format. * Generate text embeddings to reduce the VRAM bottleneck * Generate...

MarcusLoppe

Mesh conditioning instead of text conditioning

6

I was wondering if this was discussed before. The idea is to condition on existing meshes rather than text. This would be particularly useful in training it to retopologize existing...

pathquester

In the original paper, meshGPT showed it's ability to generate novel meshes based on the trained encoder-transformer models. Is it possible that a training-guide that can fulfill this function be...

HenryLhc

Loss=nan when training transformertrainer

5

I used the codes in the jupyter notebook provided by @MarcusLoppe in the discussion section, and have successfully succeeded trained the autoencoder with a loss of 0.6. However, when I...

HenryLhc

Pretrained checkpoints?

2

Are there any checkpoints of this model trained on an existing dataset? Would love to test it out!

agaier

Transformer - High VRAM, context length

4

Hello again, this issue is for next year 😃 When training the transformer, I used the follow config: ``` transformer = MeshTransformer( autoencoder, dim = 128, attn_depth = 4, attn_dim_head...

MarcusLoppe

`ResidualLFQ` was successful, but `ResidualVQ` failed severely!

2

When training the MeshAutoencoder, I compared **ResidualLFQ** and **ResidualVQ**. ResidualLFQ is your default option, which can rebuild a reasonable structure. However, when I use ResidualVQ (without changing any of your...

fighting-Zhang