Xiaoyang Wu
Xiaoyang Wu
Oh, thank you for pointing out. Their server is done. I already update the download way.
Hey, so sorry for late response. Just change `feat_keys=('coord', 'color')` is data.test to `feat_keys=("color",)` will solve this issue.
Okay, we do support multiple machines and will update the launch script (train.sh) to achieve this. Here is a pre-release as a reference (use -m to specify the number of...
Thank you. Yes, "index_valid_keys" is a new feature. I will update the configs you mentioned.
Hi, could you check `nvcc -V` has the correct output?
@theElandor Thank you for helping me clarify the batch design. This is the core of real point cloud learning (as all the point clouds in the real world have an...
Hey, we also have a mechanism triggered by `mix_prob` (usually 0.8), which will mix up point clouds one by one within a batch; maybe this made you confused. https://github.com/Pointcept/Pointcept/blob/main/pointcept/datasets/utils.py#L75
Aha, for PPT, I rewrite the trainer, and actually only use ConcatDataset.datasets to initialize multiple dataloaders. So I didn't use other features of this Class. This might be a bug?
 Strange thing, the dtype is correct on my side.
Maybe, a version issue? @RayYoh @SpeedyGonzales949 Could you share your numpy version, so that I can further track the issue? Maybe the default dtype is changed