cherry picks for 1.23.0 release #25959
jywu-msft
merged
20 commits into
rel-1.23.0
from
tlwu/1.23_cherry_pick
[CUDA] Support SwiGlu in MoE and qMoE (#25530)
b3664f88
[CUDA] BF16 MoE and qMoE (#25572)
a8e1186e
Add CUDA implementation of GatherBlockQuantized operator (#25575)
a9f74a04
Add support for QMoE in CPU (#25558)
d83904bb
Update MoE and qMoE spec (#25619)
86542411
[CPU] Improve QMoE kernel (#25822)
6ca2047b
Fix MoE CPP tests (#25877)
dd32daf7
Add custom ops library_path to EP metadata (#25830)
581b8e74
[Fix] illegal memory access in GetInputIndices with optional inputs (…
a9308a16
[TRT RTX EP] Add sync method (#25898)
6c7f150d
[TRT RTX EP] Memory map the engine buffer (#25909)
535fcc62
[TRT RTX EP] Add support for RTX runtime caches (#25917)
1f4e581a
Compile API: disable optimizations by default (#25474)
9732a3e5
[CXX] Introduce C++ API for new C entry points (#25897)
df25f456
Migrate model tests to ONNX Model ZOO only (#25888)
8f587b13
Remove std::string::data() non-const usage from public headers (#25943)
ab71f1e1
Compile API: output model and initializer stream write functions (#25…
2d36f04b
[TRT RTX EP] Fixing the stream parameter in CopyTensors API and passi…
c5096d9b
[MLAS] Add 8-bit weights ARM64 Gemm implementation (#25110)
5ee309eb
[NV TensorRT RTX] Handle unsupported data types (#25953)
157df9c6
adrianlizarraga
approved these changes
on 2025-09-05
jywu-msft
approved these changes
on 2025-09-05
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub