Christopher Lee

Hi! That's the right idea. There isn't anything built in to do that sort of thing. I suggest you start with the [forward script](https://github.com/torch-points3d/torch-points3d/blob/master/forward_scripts/forward.py) and modify it to meet your...

@ptoews Hi, we will be updating the requirements and move towards using anaconda instead of poetry in the near future. In the mean time, you should be able to install...

Hi! The S3DIS1x1 dataset comes directly from pytorch geometric, you can see information about it [here](https://pytorch-geometric.readthedocs.io/en/latest/modules/datasets.html#torch_geometric.datasets.S3DIS). What I would suggest is doing your training on the Fused dataset. If you...

Hi, that's very interesting. I have not seen that before. What model in particular are you running?

Interesting. @rasmuskongskov can you try the precompute_multi_scale suggestion and see if it fixes your issue?

I'm afraid I don't develop on windows so I'm not able to give you specific advice. However, any modern version of pytorch should be working for you, for example, 1.9.1...

Hi, you will have to do a `docker run` command with the `-p 8888:8888` option, and run jupyter with the `-i 0.0.0.0` option. I suggest you take a look here...

Hi, I'm sorry to hear you all are having issues, I will try to take a look soon! I recommend installing the master branch within a conda environment containing pytorch.

Hi! Sorry for the late response. Your dataset code looks okay to me, without digging into it too deeply. You'll want to add some transforms for your dataset with a...

Hi, I would verify that the hyperparameters in the training (scannet_benchmark/minkowski) config are the same as those used in the original paper, especially the optimizer and learning rate. Also, you...