multiple GPU

[wujunyi627]

Hi, Zheng , Thanks a lot for your work.
I was a little confused about the code: even using' --gpu_ids =0,1' or --gpu_ids=0,1,2,3, it only uses one GPU.

[layumi]

Hi @wujunyi627
For the time being, I unable the multi-gpu usage.
You may refer to this issue to change the code. layumi/Person_reID_baseline_pytorch#23 (comment)

I will definitely to merge the multi-gpu usage elegantly with the current code.

[wujunyi627]

I just made some simple modification in train_new.py.
I look forward to seeing that you definitely to merge the multi-gpu usage elegantly with the current code.

[layumi]

Thank you. But it may raise a question in test.py.

[wujunyi627]

I changed the function "save_network" in train_new.py

[nmll]

Will there be a drop when training ReID triplet loss with multi-GPU compared to the single GPU training?