Skip to content

[PyTorch] Bump minimum cuDNN version for fused attention with FP8 current scaling #2236

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

Merged
merged 3 commits into from
Oct 14, 2025
+7 −7
Merged

[PyTorch] Bump minimum cuDNN version for fused attention with FP8 current scaling #2236

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
b3eb60a
Require cuDNN 9.14.0+ for fused attention with FP8 current scaling
timmoon10 Oct 4, 2025
252be72
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Oct 4, 2025
3d383e4
Merge branch 'main' into check-cudnn-version-for-fp8_cs-attn
ksivaman Oct 9, 2025
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
14 changes: 7 additions & 7 deletions transformer_engine/pytorch/attention/dot_product_attention/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -469,13 +469,13 @@ def get_attention_backend(
fp8_recipe = fp8_meta["recipe"]
if fp8_meta.get("local_recipes", None) is not None:
fp8_recipe = fp8_meta["local_recipes"][0]
if (
use_fused_attention
and fp8_recipe.float8_current_scaling()
and device_compute_capability < (10, 0)
):
logger.debug("Disabling FusedAttention for FP8 current scaling on arch < sm100")
use_fused_attention = False
if use_fused_attention and fp8_recipe.float8_current_scaling():
if device_compute_capability < (10, 0):
logger.debug("Disabling FusedAttention for FP8 current scaling on arch < sm100")
use_fused_attention = False
elif cudnn_version < (9, 14, 0):
logger.debug("Disabling FusedAttention for FP8 current scaling with cuDNN < 9.14.0")
use_fused_attention = False

# Filter: KV cache
# backend | precision | KV cache | architecture | qkv_format | page_size
Expand Down
Loading