Open WYYAHYT opened 7 months ago
refer to dev/1.0/mmdet3d/models/roi_heads/point_rcnn_roi_head/line213, I simply add decrator @torch.no_grad()
for function _assign_and_sample()
, and the GPU memory usage remains stable during training.
refer to dev/1.0/mmdet3d/models/roi_heads/point_rcnn_roi_head/line213, I simply add decrator
@torch.no_grad()
for function_assign_and_sample()
, and the GPU memory usage remains stable during training.
Thank you very much. This solved my problem.
BTW, have you met a problem that the ETA increases instead of decreasing during the training?
Prerequisite
Task
I'm using the official example scripts/configs for the officially supported tasks/models/datasets.
Branch
main branch https://github.com/open-mmlab/mmdetection3d
Environment
Reproduces the problem - code sample
see command
Reproduces the problem - command or script
Reproduces the problem - error message
Additional information
Train Point R-CNN with KITTI dataset, it was a bug that had been fixed 2 years ago(refer #1907 and PR#1928), but it seemd that the major releases still have such bug.