There are some difficulties in running the code

[HappyBruce1]

I apologize for bothering you again. I am a beginner in the field of deep learning.I am using the train.sh script for training（only modified the path to the dataset.）, but my 4 GPUs are only able to activate one at a time. I added the --nproc_per_node=4 argument, and the code throws an error indicating that the tensors are not on the same GPU device. I am unsure whether I need to change the source code or other parameters to enable all the GPUs on my device.

[MelosY]

python -m torch.distributed.launch --nproc_per_node=4 main_finetune.py you can this script to set multi-GPU training. As your error, you can refer to https://github.com/facebookresearch/ConvNeXt-V2/blob/main/main_finetune.py.