多卡训练
CUDA_VISIBLE_DEVICES=3,6,7 python -m torch.distributed.launch --nproc_per_node=3 train.py --cfg_file cfgs/kitti_models/pv_rcnn.yaml --launcher pytorch
评测模型
python test.py --cfg_file cfgs/kitti_models/pv_rcnn.yaml --ckpt ../output/kitti_models/pv_rcnn/default/ckpt/checkpoint_epoch_80.pth