Hi, I am trying QAT for YOLOv5 and this is the 1st portion of the model
Using:
torch.quantization.fuse_modules(module, [["conv","bn","relu"]],inplace=True)
the model does not seemed fused
Whereas if i use:
torch.quantization.fuse_modules(module, [["conv","bn"]],inplace=True)
the model is fused
Is this a bug?