Skip to content

Use new sdpa_kernel API with fallback #25

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
MagellaX merged 3 commits into from
Aug 24, 2025
Merged

Use new sdpa_kernel API with fallback #25

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
096078d
Use torch.nn.attention.sdpa_kernel with fallback
MagellaX Aug 24, 2025
9f3db4e
Improve error handling and lazy imports in stream attention modules
cursoragent Aug 24, 2025
837e5c7
Merge branch 'main' into codex/update-flashattention_v3.py-for-sdpbac…
MagellaX Aug 24, 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
15 changes: 7 additions & 8 deletions stream_attention/core/flashattention_v3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,15 +80,14 @@ def forward(
if _use_flash_sdpa() and q.device.type == "cuda":
try:
# Prefer the newer torch.nn.attention API when available
sdpa_ctx = torch.nn.attention.sdpa_kernel(SDPBackend.FLASH_ATTENTION)
sdpa_ctx = torch.nn.attention.sdpa_kernel(
SDPBackend.FLASH_ATTENTION
)
except (AttributeError, TypeError):
try:
# Older PyTorch versions expose the context in torch.backends
sdpa_ctx = torch.backends.cuda.sdp_kernel(
enable_math=False, enable_flash=True, enable_mem_efficient=False
)
except (AttributeError, RuntimeError):
sdpa_ctx = nullcontext()
# Fallback for older PyTorch releases
sdpa_ctx = torch.backends.cuda.sdp_kernel(
Copy link
Contributor

@cubic-dev-ai cubic-dev-ai bot Aug 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Calling torch.backends.cuda.sdp_kernel without its own error handling can raise and crash on environments where the kernel is absent, removing the prior graceful fallback.

Prompt for AI agents 
Address the following comment on stream_attention/core/flashattention_v3.py at line 88:

<comment>Calling torch.backends.cuda.sdp_kernel without its own error handling can raise and crash on environments where the kernel is absent, removing the prior graceful fallback.</comment>

<file context>
@@ -80,15 +80,14 @@ def forward(
         if _use_flash_sdpa() and q.device.type == &quot;cuda&quot;:
             try:
                 # Prefer the newer torch.nn.attention API when available
-                sdpa_ctx = torch.nn.attention.sdpa_kernel(SDPBackend.FLASH_ATTENTION)
+                sdpa_ctx = torch.nn.attention.sdpa_kernel(
+                    SDPBackend.FLASH_ATTENTION
+                )
             except (AttributeError, TypeError):
-                try:
</file context>

enable_math=False, enable_flash=True, enable_mem_efficient=False
)

with sdpa_ctx:
out = F.scaled_dot_product_attention(
Expand Down