I use the command to start training: CUDA_VISIBLE_DEVICES=0,1 python train_s3dis.py --config config/S3DIS_Area5_3D_WSIS.yaml --pretrain log/S3DIS_Area5_3D_WSIS/epoch_00400_semantic.pth --num-gpus 2, but it shows that RuntimeError: Default process group has not been initialized.
So I added the torch.distributed.init_process_group(backend='nccl'), but is still not working.
I use the command to start training:
CUDA_VISIBLE_DEVICES=0,1 python train_s3dis.py --config config/S3DIS_Area5_3D_WSIS.yaml --pretrain log/S3DIS_Area5_3D_WSIS/epoch_00400_semantic.pth --num-gpus 2
, but it shows thatRuntimeError: Default process group has not been initialized
. So I added thetorch.distributed.init_process_group(backend='nccl')
, but is still not working.