yukang

Small gap between the reproduced and the reported results

1

Dear authors, Thanks for your good work. I have reproduced the FSA results on KITTI val split with 40 recall positions across PointPillars, SECOND, Point-RCNN, and PV-RCNN. I use 4...

Resume training from a checkpoint

14

Thanks for your great work. I implement the resuming choice for training. It can be used as following. For example, `torchrun --nnodes=1 --nproc_per_node=8 train.py --model DiT-XL/2 --data-path /path/to/imagenet/train --resume results/000/checkpoints/0100000.pt`

Hi @traveller59 , Thanks for your excellent code. We find that [focal loss](https://github.com/dvlab-research/VoxelNeXt/blob/6b6d06c4b01c8fa49b4d0d13e79ec52e89f46170/pcdet/utils/loss_utils.py#L303) fails to converge on spconv 2.2/2.3. I have tested on all versions from 2.2.2 to the newest....