Skip to content

[FLA] Add explicit kernel arguments to kda.py for Kimi Linear support #14561

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
Fridge003 merged 2 commits into from
Dec 7, 2025
Merged

[FLA] Add explicit kernel arguments to kda.py for Kimi Linear support #14561

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
166f16e
[FLA] Add explicit kernel arguments to kda.py for Kimi Linear support
alisonshao Dec 7, 2025
f47e99a
Merge branch 'main' into fix-kda-kernel-args
alisonshao Dec 7, 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
3 changes: 3 additions & 0 deletions python/sglang/srt/layers/attention/fla/kda.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,8 +102,11 @@ def fused_recurrent_kda_fwd(
# stride_final_state_token=stride_final_state_token,
# stride_indices_seq=stride_indices_seq,
# stride_indices_tok=stride_indices_tok,
USE_INITIAL_STATE=initial_state is not None,
STORE_FINAL_STATE=final_state is not None,
Comment on lines +105 to +106
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Dec 7, 2025

Choose a reason for hiding this comment

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

medium

Adding these explicit flags is a good improvement for clarity and correctness.

However, this change highlights that initial_state (and consequently final_state) can be None. The surrounding function fused_recurrent_kda_fwd does not seem to handle this case correctly and will raise an AttributeError in several places if initial_state is None:

  • initial_state.dtype on line 66.
  • initial_state.stride(0) on line 68.
  • final_state.stride(0) on line 69.

Since the variables computed on lines 68-69 are not used in the kernel call, they could potentially be removed. The issue on line 66 would also need to be addressed.

This appears to be a pre-existing issue, but it's worth fixing to prevent potential runtime errors.

IS_BETA_HEADWISE=beta.ndim == v.ndim,
USE_QK_L2NORM_IN_KERNEL=use_qk_l2norm_in_kernel,
IS_VARLEN=cu_seqlens is not None,
# INPLACE_FINAL_STATE=inplace_final_state,
IS_KDA=True,
num_warps=num_warps,
Expand Down
Loading