AssertionError assert num_gpu <= torch.cuda.device_count()

[iTruffle]

Thank you for your great job. I run the code on "Quadro RTX 5000", and successfully get the results of video demo and MOT17 MOT20 datasets. But, I fail to train on MOT17 and MOT20 datasets. These tracking and training are carried on one environment. When I train, I get an error like this:

And I have checked the informations:

How can I solve this problem? Looking forward to your reply?

[iTruffle]

I can successfully run the video demo, but fail in training.

[parthmalpathak]

Try changing - d argument to 0 or 1. It will probably work.

Otherwise try reducing the batch size as well but - d would solve it

[iTruffle]

Thank you very much. I have solved this problem. But when running the code "python3 tools/train.py -f exps/example/mot/yolox_x_mix_mot20_ch.py -d 0 -b 48 --fp16 -o -c pretrained/yolox_x.pth", I meet another error like this: How can I solve this problem?

[gyh420]

Thank you very much. I have solved this problem. But when running the code "python3 tools/train.py -f exps/example/mot/yolox_x_mix_mot20_ch.py -d 0 -b 48 --fp16 -o -c pretrained/yolox_x.pth", I meet another error like this: How can I solve this problem?

l have the same question you, how you fix it finally?

[iTruffle]

This may be the cuda version is not suitable. I run with python 3.8 torch 1.11.0+cu113 and torchvision==0.12.0

[gyh420]

This may be the cuda version is not suitable. I run with python 3.8 torch 1.11.0+cu113 and torchvision==0.12.0 thank you, l have fix it