Vchitect / Latte

Latte: Latent Diffusion Transformer for Video Generation.
Apache License 2.0
1.44k stars 147 forks source link

Questions about the *0.18215 and /0.18215 operation #72

Closed haibao-yu closed 2 months ago

haibao-yu commented 2 months ago

Hi, thanks for your excellent work. I wonder to know the meaning of mul_(0.18215) and if we can remove this operation in following command https://github.com/Vchitect/Latte/blob/c456dff74150e5b0db305fdd86b6f6de155c7634/train.py#L208

Looking forward to your reply.

maxin-cn commented 2 months ago

Hi, thanks for your excellent work. I wonder to know the meaning of mul_(0.18215) and if we can remove this operation in following command

https://github.com/Vchitect/Latte/blob/c456dff74150e5b0db305fdd86b6f6de155c7634/train.py#L208

Looking forward to your reply.

Hi, thanks for your interest. This operation (* and / 0.18215) inherits from stable diffusion. Removing this operation may result in poor results. You can refer to https://github.com/huggingface/diffusers/issues/726.

haibao-yu commented 2 months ago

Hi, thanks for your excellent work. I wonder to know the meaning of mul_(0.18215) and if we can remove this operation in following command https://github.com/Vchitect/Latte/blob/c456dff74150e5b0db305fdd86b6f6de155c7634/train.py#L208

Looking forward to your reply.

Hi, thanks for your interest. This operation (* and / 0.18215) inherits from stable diffusion. Removing this operation may result in poor results. You can refer to huggingface/diffusers#726.

Thanks for your prompt reply.