Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Lora loading for bf16 and fp8, as separate models #24

Merged
merged 12 commits into from
Oct 31, 2024
Merged

Lora loading for bf16 and fp8, as separate models #24

merged 12 commits into from
Oct 31, 2024

Commits on Sep 27, 2024

  1. Lora loading for bf16 and fp8, as separate models 

    Had to fix some bugs in the original lora loading code.

Outputs are here: https://replicate.com/replicate-internal/test-flux-dev-lora
* bf16 lora loading works
* For fp8 lora loading, the lora strength is much lower and I need to crank up lora_scale to 1.7 to match lora_scale 1.1 at bf16
    @andreasjansson
    andreasjansson committed Sep 27, 2024
    Configuration menu
    Copy the full SHA
    18553a9 View commit details
    Browse the repository at this point in the history

Commits on Oct 16, 2024

  1. Avoid weight degradation by cloning linear fp8 layers

    @andreasjansson
    andreasjansson committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    0fb3a6f View commit details
    Browse the repository at this point in the history

Commits on Oct 25, 2024

  1. handling kohya and xlabs loras

    @daanelson
    daanelson committed Oct 25, 2024
    Configuration menu
    Copy the full SHA
    538c70c View commit details
    Browse the repository at this point in the history

Commits on Oct 28, 2024

  1. don't reload/unload unnecessarily

    @daanelson
    daanelson committed Oct 28, 2024
    Configuration menu
    Copy the full SHA
    189d6ee View commit details
    Browse the repository at this point in the history

  2. small tweaks

    @daanelson
    daanelson committed Oct 28, 2024
    Configuration menu
    Copy the full SHA
    f3339be View commit details
    Browse the repository at this point in the history

  3. unified lora loading with fewer bugs

    @daanelson
    daanelson committed Oct 28, 2024
    Configuration menu
    Copy the full SHA
    bb15033 View commit details
    Browse the repository at this point in the history

Commits on Oct 30, 2024

  1. 1.5x lora scale multiplier for going fast

    @daanelson
    daanelson committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    bf910cc View commit details
    Browse the repository at this point in the history

Commits on Oct 31, 2024

  1. cog safe push ci

    @daanelson
    daanelson committed Oct 31, 2024
    Configuration menu
    Copy the full SHA
    d88d6dc View commit details
    Browse the repository at this point in the history

  2. a computer lied to me

    @daanelson
    daanelson committed Oct 31, 2024
    Configuration menu
    Copy the full SHA
    168bf87 View commit details
    Browse the repository at this point in the history

  3. Merge branch 'main' into lora-loading

    @daanelson
    daanelson authored Oct 31, 2024
    Configuration menu
    Copy the full SHA
    554712f View commit details
    Browse the repository at this point in the history

  4. linting, formatting

    @daanelson
    daanelson committed Oct 31, 2024
    Configuration menu
    Copy the full SHA
    c8f9712 View commit details
    Browse the repository at this point in the history

  5. cleanup

    @daanelson
    daanelson committed Oct 31, 2024
    Configuration menu
    Copy the full SHA
    017e384 View commit details
    Browse the repository at this point in the history