Rapid Increase in Point Cloud

[Beilinlin]

Hi！ During the training process, the initial number of point clouds was 30,000 for the first 500 iterations. However, by the 600th iteration, the number of point clouds suddenly increased from 30,000 to 239,453, and by the 700th iteration, I encountered a CUDA out of memory error. I suspect this is related to the drastic increase in the number of point clouds. Have you encountered this issue before? How can I solve this problem? Thank you!