Hi Everyone,
Can anyone suggest reasons for this kind of error and possible solution:
raise AssertionError(“Invalid device id”)
AssertionError: Invalid device id
Hi,
The problem is that the current cuda device id is not valid.
What does torch.cuda.device_count() returns? If it returns 0 that means that no device are available.
1 Like
Thank you so much for reply!
My GPU is working, I am trying this project on my personal computer (which has one GPU) as well as with Server’s GPUs.
1 Like
But what does torch.cuda.device_count() returns? You need a recent enough nvidia gpu for them to be usable.
On my system it returns 1 and on Server it is 8.
1 Like
What are the device ids you give to your DataParallel?
DataParallel is written as:
model = torch.nn.DataParallel(model, device_ids=args.gpus).cuda()
cudnn.benchmark = True
What does args.gpus contains? In particular for your system with a single gpu, it should only contain [0] and so no DataParallel should be used in that case.
My dataset is Videos so I am running on server which has to use at least 4 gpus.
if args.gpus is not None:
devices = [args.gpus[i] for i in range(args.workers)]
else:
devices = list(range(args.workers))