Skip to content

mag_bert：Train loss NaN #16

Open
Open
mag_bert：Train loss NaN#16

Description

@Jerry-xie-22
Jerry-xie-22
opened on May 7, 2025

您好，我按照readme指引运行代码，在运行text方法时，是可以正常训练和测试的。但是发现在运行mag_bert方法时，CTC对齐以后输出的视频和语音特征的Tensor值大部分为NaN，训练时loss一直为NaN，请问有解决办法嘛？

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