Can this code run on multiple GPUs?

Question

Can this code run on multiple GPUs?

lxc86739795 opened this issue 5 years ago · comments

Xinchen Liu commented 5 years ago

It seems that the code cannot directly run on multiple GPUs by simply adding "torch.nn.Dataparallel".

Artsiom · Answer 1 · Thu Jul 04 2019 23:58:59 GMT+0800 (China Standard Time)

I'm not sure about that. Maybe you will have to make some extra modifications.
What kind of error do you get?

Chen94yue · Answer 2 · Fri Jul 05 2019 10:58:48 GMT+0800 (China Standard Time)

LOL @lxc86739795

Faiz Ul Wahab · Answer 3 · Sun Jul 07 2019 02:11:55 GMT+0800 (China Standard Time)

The simplest way to go about this is to make modifications until train_batch and disable faiss_gpu. You might have to make more modification in terms of gpu sharing between faiss and the training session.