Add MHA fusion for Nemotron speech conformer encoder #27764
kunal-vaishnavi
merged
11 commits into
microsoft:main
from
kunal-vaishnavi:kvaishnavi/nemotron-conformer
Add MHA fusion for conformer encoder
bf02d86b
tianleiwu
approved these changes
on 2026-03-19
tianleiwu
requested changes
on 2026-03-19
Fix conformer attention fusion regression and redundant postprocess c…
62e6357d
Add unit tests for new conformer attention graph pattern matchings
ff20428a
Fix conformer test models: use non-zero qk_bias to survive ORT consta…
673bbc76
Fix RUFF-FORMAT linting: collapse multi-line numpy_helper.from_array …
fd2698a8
Merge branch 'copilot/address-issues-from-review' into kvaishnavi/nem…
4c8305de
tianleiwu
requested changes
on 2026-03-24
Refactor extra_q_nodes condition for clarity
0bf04834
Remove postprocess method from onnx_model_conformer
305a2aff
Cherry pick Copilot fix: Fix attention_bias 4D shape validation in Fu…
2a8630bb
tianleiwu
commented
on 2026-03-25
tianleiwu
dismissed these changes
on 2026-03-25
Clarify attention_bias shape requirements in comments
bb5f7f4d
github-advanced-security
commented
on 2026-03-25
Remove trailing whitespace
a707d9e2
tianleiwu
approved these changes
on 2026-03-27
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub