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

Update README.md for float8 unification #895

Merged
merged 1 commit into from
Sep 16, 2024
Merged

Update README.md for float8 unification #895

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions torchao/float8/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,8 @@ throughput speedups of up to 1.5x on 128 GPU LLaMa 3 70B pretraining jobs.

:warning: <em>The codebase is stable, but backwards compatibility is not yet guaranteed.</em>

:warning: <em>These APIs are training-only and float8-only, and we plan to [unify them with the rest of torchao](https://github.com/pytorch/ao/issues/894) in the future.</em>

# Single GPU User API

We provide three per-tensor scaling strategies: dynamic, delayed and static. See https://arxiv.org/pdf/2209.05433.pdf, Section 4.3 for more details. These strategies are configurable separately for activations (`input`), weights (`weight`) and gradients (`grad_output`).
Expand Down
Loading