Jiacheng Liu

> This error happens because the number of points per batch is too big or too small. Their should be plots appearing when this error happens, could you show them...

The first photo shows the dataset which I have modified type as s3dis

> Also I assume your dataset has varying densities like Semantic3D for example As you assumed this dataset is a mobile LiDAR dataset too, so I will modify this code...

After my attempts, I modified `in_radius` and `first_subsampling_dl` as the photo shows, this trainer runed successfully, but few minutes later, this error raised: I assume this error is happening in...

Is this because I modified the .ply file of "Toronto" dataset directly instead of pre-processing it as a .txt file? The original format of this dataset is .ply files, so...

> In `trainer.py` just before line 539, try to print the shape of `truth` and `preds`, if both are [0], then it means you have an empty input sphere I...

Thank you for your patient reply!!I even suspected that it was caused by too much difference between this data and s3dis, I had also considered switching to the kitti form,...

Hello, Mr.Thomas! I tried to place these codes below the ``` predictions.append (probs) targets.append (target) 10 += length ``` like the photo shows, but it printed nothing, and I deleted...

Thank you for your reply! I changed it but it still did not work, so I deleted the "if statement". The print message is as the photo shows: ...

@SC-shendazt Thank you!! Actually i am trying to do this, but the gpu_memory of this computer is not large enough. so i have to try it on a more powerful...