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

[feat] Dropout partial bw fusion (second take) #164

Merged
merged 7 commits into from
Jan 3, 2022
Merged

[feat] Dropout partial bw fusion (second take) #164

merged 7 commits into from
Jan 3, 2022

Commits on Dec 23, 2021

  1. flipping the seeds so that it drops down from the top 

    using less seeds

tiling + vertical seeds

Computing the FW and BW per tile over M

better scheduling defaults, improves across the board

good enough perfs

catching the slow case and diverting to pytorch in that case
    @blefaudeux
    blefaudeux committed Dec 23, 2021
    Configuration menu
    Copy the full SHA
    b0d5f91 View commit details
    Browse the repository at this point in the history

  2. Fixing the randomness problem, updating the graphs

    @blefaudeux
    blefaudeux committed Dec 23, 2021
    Configuration menu
    Copy the full SHA
    9ee09ee View commit details
    Browse the repository at this point in the history

Commits on Dec 26, 2021

  1. WIP better scheduling, improving perfs all around

    @blefaudeux
    blefaudeux committed Dec 26, 2021
    Configuration menu
    Copy the full SHA
    b36e8ca View commit details
    Browse the repository at this point in the history

Commits on Dec 27, 2021

  1. larger tiles and better speed, full loss curve checked

    @blefaudeux
    blefaudeux committed Dec 27, 2021
    Configuration menu
    Copy the full SHA
    8b9befa View commit details
    Browse the repository at this point in the history

Commits on Jan 3, 2022

  1. code review, not all factorization options were possible in the end, … 

    …Triton jit can be a little picky
    @blefaudeux
    blefaudeux committed Jan 3, 2022
    Configuration menu
    Copy the full SHA
    dee2fe8 View commit details
    Browse the repository at this point in the history

  2. tweaking formulations so that they please Triton. Slightly faster dro… 

    …pmask generation
    @blefaudeux
    blefaudeux committed Jan 3, 2022
    Configuration menu
    Copy the full SHA
    3e74a7e View commit details
    Browse the repository at this point in the history

  3. small cleanup, removing old files and fixing broken links

    @blefaudeux
    blefaudeux committed Jan 3, 2022
    Configuration menu
    Copy the full SHA
    c40dd64 View commit details
    Browse the repository at this point in the history