liangleikun
Results
2
comments of
liangleikun
Hello, I have the same problem. I used LRS2 dataset to train lip_sync_net, but the loss does not go down. It is 0.68. Could you give me some advice?
> 试试把seed设置成0 还是一样有差异，而且应该和seed没关系，不是训练阶段，是模型eval()推理的时候，理论上推理的时候不应该有某个随机结果，没想通差异是为啥