Skip to content

[enhancement]: Add support for SVDQuant #7933

Open
@xroberx

Description

@xroberx

Is there an existing issue for this?

  • I have searched the existing issues

Contact Details

No response

What should this feature add?

Hi,

SVDQuant allows to use FLUX on low end hardware by reducing memory usage and increasing performance. For example, with my RTX 3050 (8GB), the iteration time goes from 6 seconds per iteration to 1.5 seconds, eliminating the need to keep using SDXL (at least in my case). At the moment, the project is completely usable even though they are adding new features and fixing some minor bugs. In fact, I am using it within ComfyUI, but I would really love to be able to use it within InvokeAI, because the interface is so much better.

References:
https://github.com/mit-han-lab/nunchaku/tree/main
https://github.com/mit-han-lab/ComfyUI-nunchaku

Best regards,
Robert.

Alternatives

No response

Additional Content

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions