Skip to content

音频特征缩放参数 audio_scale #71

Open

Description

@wangjue-wzq

代码中直接通过 audio_scale 进行特征融合的缩放,推理的时候缩放越大口型越大,请问训练的时候是怎么做的?论文中有提到训练的时候以一定概率置 0,但是对于缩放的系数可以设置比如4、5、6 这么大的值,还不会明显变差,请问为什么?

https://github.com/Fantasy-AMAP/fantasy-talking/blob/main/model.py#L86

        x = x + img_x + audio_x * audio_scale
        x = attn.o(x)
        return x

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions