Open miganchuanbo opened 1 year ago
https://github.com/lucidrains/DALLE2-pytorch/blob/c3df46e3743eccd951ac71eb1d476910ace5aabc/dalle2_pytorch/dalle2_pytorch.py#L2076
Why do we need to zero init the conv layer?
https://github.com/lucidrains/DALLE2-pytorch/blob/c3df46e3743eccd951ac71eb1d476910ace5aabc/dalle2_pytorch/dalle2_pytorch.py#L2076
Why do we need to zero init the conv layer?