microsoft/onnxruntime

Pull Requests Commits

Copilot committed 23 days ago

Verified 0ae975a0

Replace old use of `routing_record` with `routing_snapshot_` in onnxruntime/contrib_ops/cuda/moe/moe_quantization.cc. (#32932)

edgchen1 committed 23 days ago

Verified 80291e14

[CUDA] Replace shared ones-buffer bias broadcasts in Gemm and DecoderAttention (#32852)

Copilot committed 23 days ago

Verified 8ba5bec8

[CUDA] Auto-tune small-M decode MatMul (cuBLAS vs small-N GEMV) and speed up MatMulNBits GEMV (#32876)

tianleiwu committed 23 days ago

Verified 567deb38

[WebGPU] Fix initializer upload on GPUs without Resizable BAR (#32816)

SharmaRithik committed 24 days ago

Verified 96f73115

Expose CoreML EP via OrtEpFactory (#31975)

maxim-davgalev committed 24 days ago

Verified 626a7c56

Update MoE offload implementation plan and implements a global state for MoE/QMoE (#32738)

xadupre committed 24 days ago

Verified c2650ccb

[WebGPU] Keep LayerNormalization arithmetic in FP32 (#32896)

gyagp committed 24 days ago

Verified e927db3f

Use WRL COM pointers for DXCore device discovery (#32815)

mustjab committed 24 days ago

Verified 78c4614f

Reduce coordinate work in CPU N-D Im2col (#32817)

Hosi121 committed 24 days ago

Verified 38aee156

[MLAS] Avoid unused dynamic QGEMM packed buffer allocation (#32879)

mirounga committed 24 days ago

Verified d194fe39

Local non-causal for paged attention (#32792)

aciddelgado committed 24 days ago

Verified e1805779

Add github action for arm64 windows build (#32791)

eserscor committed 24 days ago

Verified e758c960

Validate sparse initializer value cardinality (#32753)

apsonawane committed 24 days ago

Verified 2c8379bc

Fix the nlohmann_json tarball hash in the standalone model_package build (#32596)

rvandermeulen committed 24 days ago

Verified 84cf511a

Use onnxruntime::narrow instead of gsl::narrow in core framework/graph/optimizer (#32591)

rvandermeulen committed 24 days ago

Verified 6b211684

Use ORT_TRY/ORT_CATCH for the remaining raw try/catch in the core non-minimal build (#32597)

rvandermeulen committed 24 days ago

Verified 20ba75f2

Fix Clang portability in logging, GELU, and spin pause (#32853)

mustjab committed 24 days ago

Verified 35597c6f

[WebGPU] Support concurrent use with explicit Session streams (#32587)

xiaofeihan1 committed 25 days ago

Verified 2b758bc6

Add packed INT2 and mixed-width CUDA QMoE decode (#32761)

jiafatom committed 25 days ago

Verified a11b4e59

Fix file placement for common linear attention file (#32831)

kunal-vaishnavi committed 25 days ago

Verified 92594005

test fix implemented (#32798)

melkap01-Arm committed 26 days ago

Verified cfad0a05

[CUDA] MatMulNBits: optional M chunking for the fpA_intB path (#32810)

tianleiwu committed 26 days ago

Verified 3756d4dc

[CUDA] Give each plugin EP session its own device arena (#32807)

tianleiwu committed 27 days ago

Verified 4c868628

Fix WebGPU build error when LTO is enabled (#32813)

skottmckay committed 27 days ago

Verified d4bd8135

Fix Python API docs ONNX dependency (#32811)

damdoo001-arm committed 27 days ago

Verified 3482ff20

Validate dynamic pads input access (#32654)

apsonawane committed 27 days ago

Verified f1fae4c9

Fuse BatchNormalization into ConvTranspose (#29542)

the0cp committed 27 days ago

Verified b61f007b

Enable non-causal WebGPU PagedAttention for full-context blocks (#32763)

aciddelgado committed 27 days ago

Verified 83e48928

WebGPU: add physical Dawn adapter index selection (#32319)

tianleiwu committed 27 days ago

Verified d306e3ee

Older