Add support for `candle-flash-attn` and `candle-flash-attn-v3` in CUDA - #14

Merged

alvarobartt merged 6 commits into

add-missing-laya-variantsfrom

add-flash-attention

Sep 25, 2026

Commits on Sep 25, 2026

Add support for `candle-flash-attn` and `candle-flash-attn-v3`
alvarobartt
committed
774ab2d
Update `README.md`
alvarobartt
committed
489d9d2
Update `.github/workflows/ci.yml`
alvarobartt
committed
c13bb3a
Use `match` over `if` when applicable on attention
alvarobartt
committed
e2e408e
Add `build.rs` in Dockerfiles & pre-define features based on compute cap
alvarobartt
committed
db79e80
Use `flash-attn-2` for Hopper in `Dockerfile.cuda`
alvarobartt
committed
2e821f3