Feng Wang
请确认是不是加入的是2群，1群人已经满了。所有加入1群的请求都无法通过。
One anchor matches one box in a single class.
One box multi-classes is widely used, e.g. RetinaNet, FCOS(acnhor free). In my opinion, the method you are talking is used in two stage detector like Faster R-CNN.
From your log: ``` import torch torch.backends.cuda.matmul.allow_tf32 = True torch.backends.cudnn.benchmark = True torch.backends.cudnn.deterministic = False torch.backends.cudnn.allow_tf32 = True data = torch.randn([4, 128, 20, 20], dtype=torch.half, device='cuda', requires_grad=True) net = torch.nn.Conv2d(128,...
NCCL is not supported by windows, plz use gloo instead.
Thanks for your contribution @daquexian , Plz lint your code to pass the workflow. BTW, have your ever double checked exported onnx model ?
> @FateScript Fixed. > > Could you please help check it if there is some concern? Ok, I will check this later. Plz remind me if this PR is not...
Could you plz show me some error log @firasjlassi @NourSoltani ?
The true reason here is `thop` not onnx, try downgrade the package.
Did you modify any code?