Open gcunhase opened 1 year ago
torchrun --nproc_per_node=4 main.py --eval \ --cfg ./configs/swin/swin_tiny_patch4_window7_224.yaml \ --resume ../swin_tiny_patch4_window7_224.pth \ --data-path /data1/imagenet/ \ --batch-size=128
Log:
Swin Transformer training and evaluation script: error: the following arguments are required: --local_rank
Reason:
local_rank
--nproc_per_node>1
Related: https://github.com/microsoft/Swin-Transformer/issues/158
Hi, have you solved it?
Command to repro
Issue
Log:
Reason:
local_rank
arg is required, see here--nproc_per_node>1
and also setlocal_rank
as some int, only 1 GPU is used. How can I use multiple GPUs?Related: https://github.com/microsoft/Swin-Transformer/issues/158