Skip to content

realtime quantization to workers #63

Description

@evilsocket

in theory, if the master has the full res model, while streaming to workers it could quantize it automatically on the fly so that a worker can host a quantized version of the full model the master is using / orchestrating.

Activity

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

Metadata

Metadata

Assignees

Labels

enhancementNew feature or request

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions