Skip to content

Loss Nan when training the mode #1

@mlkk518

Description

@mlkk518

Thanks for sharing your wonderful work! The training goes into failure case after several training iterations, and 'Nan' is printed. Though I have set the learning rate to 1e-8, and clip the gradient, the issue still exists. Could you please help me go against the problem. Thank you very much! Best regards.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions