Skip to content

webgpu: bypass manual mRoPE for text-only Qwen3.5 when GQA fuses RoPE#2245

Draft
qjia7 wants to merge 1 commit into
mainfrom
qwen35-text-only-fused-rope-bypass
Draft

webgpu: bypass manual mRoPE for text-only Qwen3.5 when GQA fuses RoPE#2245
qjia7 wants to merge 1 commit into
mainfrom
qwen35-text-only-fused-rope-bypass

webgpu: bypass manual mRoPE for text-only Qwen3.5 when GQA can fuse RoPE

86440e3
Select commit
Loading
Failed to load commit list.
GitHub Advanced Security / CodeQL failed Jun 26, 2026 in 4s

1 configuration not found

Warning: Code scanning may not have found all the alerts introduced by this pull request, because 1 configuration present on refs/heads/main was not found:

API upload

  • ❓  <default>

New alerts in code changed by this pull request

  • 1 error
  • 1 warning

See annotations below for details.

View all branch alerts.

Annotations

Check failure on line 35 in src/python/py/models/builders/qwen.py

See this annotation in the file changed.

Code scanning / CodeQL

Wrong number of arguments in a call Error

Call to
method Model.make_attention_init
Loading
with too few arguments; should be no fewer than 1.

Check warning on line 32 in src/python/py/models/builders/qwen.py

See this annotation in the file changed.

Code scanning / CodeQL

Signature mismatch in overriding method Warning

This method requires 1 positional argument, whereas overridden
Model.make_attention_init
Loading
requires 2.
This call
Loading
correctly calls the base method, but does not match the signature of the overriding method.