training is slow

Open U019TT opened this issue 4 years ago • 0 comments

When running pre-train task on 4 V-100 GPUs, I found that this line of code in shuffle BN takes a lot of time: idx_shuffle = torch.randperm(batch_size_all).cuda()

In addition，speed of RPN head is also slow.

Do you know what's going on？ Look forward to your reply.

Dec 15 '21 03:12 U019TT