train

[twsha]

hello,i want to ask that I set multi-gpu training to single-gpu training，But report the following error：

and I set it like this in the source code：

[twsha]

How should I further set up the code to change to single gpu training

[zyxu1996]

Try this

[zyxu1996]

CUDA_VISIBLE_DEVICES=0 python -m torch.distributed.launch --nproc_per_node=1 --master_port 29506 train.py

[twsha]

I train directly through the train file，find that it can't read dateset, can you tell me that below the annotaions file The train.txt under the annotations file have content? thank you so much!

[zyxu1996]

[twsha]

thanks