Skip to content

DQ→MatMulNBits fusion transformer for NvTensorRtRtx ep #27466

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
tianleiwu merged 8 commits into from
Mar 5, 2026
Merged

DQ→MatMulNBits fusion transformer for NvTensorRtRtx ep #27466

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
7e66590
optimizer: fuse WebNN DQ chain back to MatMulNBits for NvTensorRTRTX
anujj Feb 18, 2026
0434a1f
optimizer(webnn): keep DQ->MatMulNBits fusion on phase-1 safe path
anujj Feb 19, 2026
e38ae4f
optimizer(webnn): restore Gemm bias preserving DQ to MatMulNBits fusi…
anujj Feb 19, 2026
e433249
Add SDXL Turbo fusion pattern for QDQ+MatMul -> MatMulNBits fusion
praneshgo Feb 24, 2026
887b625
optimizer(webnn): harden direct DQ->MatMulNBits fusion shape checks
anujj Feb 25, 2026
f98418f
file name changes
anujj Feb 25, 2026
8559bcd
refactor: split DQMatMulNBitsFusion::ApplyImpl into focused helpers
anujj Feb 28, 2026
40207a9
address PR review: config-driven gating, SafeInt, minimal-build guard…
anujj Mar 4, 2026
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
5 changes: 5 additions & 0 deletions include/onnxruntime/core/session/onnxruntime_session_options_config_keys.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -385,6 +385,11 @@ static const char* const kOrtSessionOptionsMlasLutGemm = "mlas.use_lut_gemm";
// If not provided, default is 4.
static const char* const kOrtSessionOptionsQDQMatMulNBitsAccuracyLevel = "session.qdq_matmulnbits_accuracy_level";

// Enable the DQ->MatMulNBits fusion graph transformer.
// "0": disabled (default). "1": enabled.
// This is typically set automatically by InferenceSession when the NvTensorRTRTX EP is registered.
static const char* const kOrtSessionOptionsEnableDQMatMulNBitsFusion = "session.enable_dq_matmulnbits_fusion";

// THIS OPTION IS NOT A REGULAR SESSION OPTION SINCE IT CAN BE MODIFIED AT ANY TIME
// Meant to be used with SetEpDynamicOptions
// Specify the type of workload for this session.
Expand Down
Loading
Loading