search

zhanglonghao1992 / One-Shot_Free-View_Neural_Talking_Head_Synthesis

Pytorch implementation of paper "One-Shot Free-View Neural Talking-Head Synthesis for Video Conferencing"
Other
735 stars 141 forks source link

出现错误：size mismatch for first.conv.weight #31

Open ruanjiyang opened 2 years ago

ruanjiyang commented 2 years ago

您好，我在尝试跑demo的时候，发现这个错误 size mismatch for first.conv.weight: copying a param with shape torch.Size([64, 3, 7, 7]) from checkpoint, the shape in current model is torch.Size([64, 3, 3, 3]).

请问如何解决呢？谢谢！

zhanglonghao1992 commented 2 years ago

@ruanjiyang If you are using checkpoints of the beta version, make sure that the --gen is set to "original", and get_rotation_matrix() corresponds to the beta version when running the demo.py. I suggest you train the new version of the model and test it.

bronzeman2022 commented 1 year ago

@ruanjiyang 大佬 你最后咋解决的 我也遇到了这个问题 按照代码作者的方式好像没有