Skip to content

Switch attention backend from flex to cuda #139

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 1 commit into from
Aug 29, 2025
Merged

Switch attention backend from flex to cuda #139

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: 1 addition & 1 deletion examples/modeling/modeling_doge.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -304,7 +304,7 @@ def forward(
attention_mask=attention_mask,
)

attention_interface: Callable = flash_dmattn_func_auto(backend="flex")
attention_interface: Callable = flash_dmattn_func_auto(backend="cuda")
Copy link

Copilot AI Aug 29, 2025

Choose a reason for hiding this comment

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

The flash_dmattn_func_auto function raises an ImportError regardless of the backend parameter, as shown in the context. This change will not have any functional effect since the function always fails. The underlying flash_dmattn dependency needs to be properly installed and the function implementation needs to be updated to actually use the backend parameter.

Copilot uses AI. Check for mistakes.
query_states = query_states.transpose(1, 2).contiguous() # [B, H, Q_LEN, D]
key_states = key_states.transpose(1, 2).contiguous() # [B, H, KV_LEN, D]
value_states = value_states.transpose(1, 2).contiguous() # [B, H, KV_LEN, D]
Expand Down