HuangJunJie2017 / BEVDet

Code base of the BEVDet series .
Apache License 2.0
1.45k stars 266 forks source link

多GPU #338

Open yang-kuo opened 7 months ago

yang-kuo commented 7 months ago

尊敬的作者您好:我想要在train.py文件里面修改参数来实现多gpu训练应该怎么改呢,我尝试了但没成功 }PCDKH$LJ(5R QZJ6 XY5MD 12$KG9H~M}RVLM2~)L`LK5P

YuhanZhou521 commented 6 months ago

尊敬的作者您好:我想要在train.py文件里面修改参数来实现多gpu训练应该怎么改呢,我尝试了但没成功 }PCDKH$LJ(5R QZJ6 XY5MD 12$KG9HM}RVLM2)L`LK5P

请问你解决了吗?

L1bertad commented 6 months ago

可以使用这个命令实现CUDA_VISIBLE_DEVICES="1" python -m torch.distributed.launch --nnodes 1 --node_rank 0 --master_addr "127.0.0.1" --nproc_per_node 1 --master_port 29500 tools/train.py configs/xxx --seed 0 --launcher pytorch 其中CUDA_VISIBLE_DEVICES设置为对应的gpu-id,nproc_per_node 设置为对应gpu数量