Xiaoyang Wu

Hi, reading and understanding our preprocessing code for ScanNet is all you need~ https://github.com/Pointcept/Pointcept/blob/main/pointcept/datasets/preprocessing/scannet/preprocess_scannet.py If you have any questions related to the detailed preprocessing, feel free to ask me.

Hi, could you provide a more detail config, as any training parameters might cause this issue?

Hi, use our schedular for other PTv3 configs. Setting the learning rate for the attention layer to 1/10 of the usual learning rate is important for PTv3. https://github.com/Pointcept/Pointcept/blob/main/configs/scannet/semseg-pt-v3m1-0-base.py#L56-L65

Hi, I think it is normal for some special criteria (loss).

> ame exp name and there exactly exist an model in ./exp/modelnet40/PTv Thanks for helping me solve the issue. Example configs for loading weight and fine-tuning is available in MSC...

Sorry, I also can not figure out the reason.

Hi, you can consider download one processed ScanNet dataset, and check the data structure.

As most of the config for test dataset, GirdSampling in the test.transform is not test mode. https://github.com/Pointcept/Pointcept/blob/main/configs/nuscenes/semseg-pt-v3m1-0-base.py#L164

> Do you think that I should split my training cloud into multiple file Typically no need for this. > will this be managed by the loader Managed by dataset...

Here is an example: https://github.com/Pointcept/Pointcept/blob/main/configs/scannet/semseg-pt-v3m1-0-base.py#L134. Here, the feat_key is ["coord", "normal"], which means that the Collect function will create "feat" with "coord" and "normal" ([n, 6]), further resulting in an...