feat: FA3-FP8-extension#552

Merged

Marius-Graml merged 16 commits intomainfrom

feat/fa3-fp8-extension

Apr 14, 2026

Commits on Mar 9, 2026

Initial commit: Add fp8 quantization to fa3 as hyperparameter
Marius-Graml
committed
Fix: ruff checks
Marius-Graml
committed
Add parameter fp8 to documentation of FlashAttention3Context class
Marius-Graml
committed
Add target modules for older and newer diffusion version
Marius-Graml
committed
Remove prints, remove fixed wrapper flag, pass ruff checks.
Marius-Graml
committed
Resolve PR comments. Instead of using del for q,k,v directly assign their quantized versions to their fp16 ones. Instead of checking for diffusers pipeline version, check for diffusers pipeline (an…
Marius-Graml
committed
Resolve merge conflicts in flash_attn3.py
Marius-Graml
committed

Commits on Mar 16, 2026

Commits on Mar 23, 2026

Commits on Mar 25, 2026

Commits on Mar 30, 2026

Add cast for target modules
Marius-Graml
committed

Commits on Apr 14, 2026

Merge branch 'main' into feat/fa3-fp8-extension
Marius-Graml
authored