Christopher Lee

@ptoews Ah, I see. I'm not too sure to be honest, but I suspect it has something to do with this: https://github.com/torch-points3d/torch-points3d/pull/353/files ` input_nc=input_nc, # KPconv is particular. Pos aren't...

Hi Diwakar, I'm guessing you're using the `forward_scripts/forward.py` script for the forward pass? Please try the following for me: In `forward_scripts/forward.py` replace `@hydra.main(config_path="conf/config.yaml")` with `@hydra.main(config_path="conf", config_name="config")`. Let me know if...

Hi, `models` is a combination of task and model_type. `data` is a combination of task and `dataset`. So: ``` conf/data/{data}.yaml conf/models/{models}.yaml ``` An example command: ``` python train.py task=segmentation models=segmentation/sparseconv3d...

@OOps717 Hi, I think this might be fixed by #643 can you take a look?

Hi, it looks like you may be using an outdated version of the code. I think you'll find the issue won't happen if you update to the latest master branch....

Hi, I think adding a docker file that supports 1.9.0+ and cuda 11.1 would be great! I don't immediately see anything that wouldn't be compatible, just make sure you upgrade...

Hi, Thanks for the suggestion! The reason the requirements.txt is so specific is because it's being generated automatically from `poetry`, however, I think it's out of sync in the repo,...

It looks like there's not a single scale encoder architecture defined. Take a look at https://github.com/nicolas-chaulet/torch-points3d/tree/master/torch_points3d/applications/conf/pointnet2 and you can see what configs there are. For a multi-scale encoder architecture: ```...

I'll take a look after this semester ends on redoing a lot of the broken pn++ things :)

That's interesting, here's where it came from: https://github.com/conda-forge/staged-recipes/pull/15095 It looks like it's CPU only builds at the moment. I will reach out to the maintainer and see if we can...