Skip to content

Question: How to use Shared LLM embeddings in stage 1 for alignment? #4

Open
@njzheng

Description

@njzheng

Since u only use CTC-loss, is there any detail about how to convert the audio features into the representation space of LLMs with a Transformer layer?
How to use existing LLM embeddings here?

Thank you!

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