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

Make fp8 work on older GPUs #34

Merged
merged 5 commits into from
Nov 4, 2024
Merged

Make fp8 work on older GPUs #34

merged 5 commits into from
Nov 4, 2024

Commits on Nov 1, 2024

  1. fp8: fall back to float32 matmul on cuda capability < 8.9 

    This re-enables the use of fp8 on older GPUs, which can be useful
to save vram.
    @yorickvP
    yorickvP committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    27181bd View commit details
    Browse the repository at this point in the history

  2. fp8: override offload/compile config based on what we're doing

    @yorickvP
    yorickvP committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    f06f157 View commit details
    Browse the repository at this point in the history

  3. Check video memory to decide when to offload

    @yorickvP
    yorickvP committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    51e7ee4 View commit details
    Browse the repository at this point in the history

  4. Ruff format

    @yorickvP
    yorickvP committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    0aac308 View commit details
    Browse the repository at this point in the history

  5. Don't compile fp8 when offloaded, it's going to be slow anyways

    @yorickvP
    yorickvP committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    4fcda4d View commit details
    Browse the repository at this point in the history