Closed ogkalu2 closed 1 year ago
If I don't specify cpu when loading the models, I get
File review A = w_a @ w_b.T "Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!"
I now know the linear sum assignment message it can't ask run on a GPU but is it possible to get it to use both ?
I'm no pytorch expert, but you can use x.cpu() to move something to the CPU, and then x.cuda() to move it back
x.cpu()
x.cuda()
yea, just move everything to gpu, data and model.
If I don't specify cpu when loading the models, I get
File review A = w_a @ w_b.T "Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!"
I now know the linear sum assignment message it can't ask run on a GPU but is it possible to get it to use both ?