meshgpt-pytorch

meshgpt-pytorch copied to clipboard

Mesh completion task

2

Hello, Mesh/shape completion application is mentionned in the paper (see Figure 9). I was wondering how to apply this kind of feature and I would have three questions: - Once...

jrmgr

Here is from line 1442 to line 1449 in ['meshgpt_pytorch.py':](https://github.com/lucidrains/meshgpt-pytorch/blob/main/meshgpt_pytorch/meshgpt_pytorch.py#L1442) ` sos = repeat(self.sos_token, 'd -> b d', b = batch) # b, nf, d attended_face_codes_with_sos, _ = pack([sos, attended_face_codes],...

LeeYan1

distirbution training

1

Hello, first of all thank you for sharing. I have a question for you. How to use your code to achieve stand-alone multi-gpu distributed training ah. Currently, with your code,...

kingofrubbish2

ImportError: cannot import name 'MeshDataset' from 'meshgpt_pytorch'

5

Hi, I am running the demo and it seems like MeshDataset can not be imported from meshgpt_pytorch. Any help @lucidrains would be greatly appreciated! ``` --------------------------------------------------------------------------- ImportError Traceback (most recent...

StephenYangjz

Sliding window for transformer

1

Hi, I was wondering if it was possible to implement a sliding window decoder for the transformer? When increasing the max sequences length, the training time goes up dramatic and...

MarcusLoppe

Question

10

Hi everyone, I had some questions about this method: - The transformer seems to learn about 3D meshes as a sequence, but the order of that sequence must have a...

nicolasdonati

Commit loss is negative

9

When I trained on several objects with several epochs, the commit loss starts to become negative, and it turns out that the overall loss keeps going down, but neither the...

ZekaiGalaxy

Simple training script for toy data?

6

Hi there, I wonder if it's possible to have some script reproducing the same toy [example](https://github.com/google-deepmind/deepmind-research/blob/master/polygen/sample-pretrained.ipynb) from an older paper. I tried to run the training, but the best thing...

xiao-xian

Data augmentation strategies

56

In https://github.com/lucidrains/meshgpt-pytorch/pull/6 For each mesh I generate augments_per_item (like 200), then I use it to index into the dataset. Using a seed I augment using this strategy. What do you...

fire

Since my other proposals didn't mention what it would take to be merged, I will ask before opening prs to avoid wasted effort. I made a change to meshgpt-pytorch https://github.com/V-Sekai-fire/meshgpt-pytorch/blob/main/run.py...

fire