microsoft/onnxruntime

Pull Requests Commits

Fix FuseInitializersTransformer consumer input lookup (#32426)

sylvesterkaczmarek committed 9 hours ago

Verified bb331b7a

Fix CUDA plugin user stream test lifetime (#32449)

Copilot committed 14 hours ago

Verified 9f913ae5

[js/web] Support float16 in Conv3DNaive WebGPU shader (#32357)

Novestars committed 16 hours ago

Verified 6690cf0d

[WebGPU] Share subgroup matrix MatMul with pointwise Conv (#32304)

qjia7 committed 3 days ago

Verified eebea690

[CUDA] Register BF16 ReduceMean kernels (#32326)

tianleiwu committed 3 days ago

Verified 3af6be47

Remove Hugging Face download dependency from test_dynamic_quantization_subgraph (#32443)

Copilot committed 3 days ago

Verified 781a3258

NetworkIsolation for NPM. Remove some unused parameters from release pipelines. Set default RC to 1. (#32440)

eserscor committed 3 days ago

Verified 208b435b

[MLAS] SVE Linear Attention Kernel (#32356)

mirounga committed 3 days ago

Verified 1756ea0d

Tune FP4 GEMV scheduling for the 48-SM SM121 GPU (#32408)

baijumeswani committed 3 days ago

Verified bae2f939

[CUDA] Add an opt-in FP8 DeepGEMM MoE decode path for QMoE (#32122)

tianleiwu committed 3 days ago

Verified aeb1a16a

[CUDA] Add speculative decoding to paged XQA (#32340)

tianleiwu committed 4 days ago

Verified 491a5f5f

Add packed attention workspace estimates (#32321)

titaiwangms committed 4 days ago

Verified 51df5b52

[CUDA] Optimize TopK for wide last axes (#32404)

tianleiwu committed 4 days ago

Verified 31ca9061

Reset plugin EP stream chunks before release (#31983)

apsonawane committed 4 days ago

Verified d47fd882

Bound cross_qk layer and head indices in CUDA Whisper beam search (#31998)

apsonawane committed 4 days ago

Verified 90e2f53b

Avoid pinned buffers in CUDA Split and Concat fast paths (#32410)

baijumeswani committed 4 days ago

Verified 2af39c97

Make workspace input shapes optional-aware (#32312)

titaiwangms committed 4 days ago

Verified f98e809f

Avoid overflow in CUDA reduction scans (#32137)

apsonawane committed 4 days ago

Verified bfa00fa6

Add AVX2 LayerNorm/RMSNorm kernel for x86-64 (MLAS) (#31973)

justinchuby committed 4 days ago

Verified 6dd9ba94

Add GetPreallocatedOutput to KernelContext. (#32089)

ericcraw committed 4 days ago

Verified e1bb372a

[WebNN] Fix inverted fallback data type support check (#32293)

Honry committed 4 days ago

Verified 66555cc4

[MLAS] Do not build the POWER MLAS kernels when VSX is unavailable (#32389)

pkubaj committed 4 days ago

Verified a535bcfe

[MLAS] Restore the run time vector extension check on riscv64 (#32406)

Zestion committed 4 days ago

Verified fb765166

Enable LayerNormFusion for the WebGPU EP to fix fp16 inference correctness (#32294)

Lapis0x0 committed 4 days ago

Verified e75f6c89

Separate onnxruntime_providers compilation from its archive (#32162)

rvandermeulen committed 4 days ago

Verified fe52a3ce

[CUDA] Fix PagedAttention build without FlashAttention (#32327)

tianleiwu committed 4 days ago

Verified 2ea96f75

[CUDA EP] Slice: single-memcpy fast path for contiguous subregions (#28902)

Copilot committed 5 days ago

Verified dcc70f5c

[TRT] Fix trt shape values (#32415)

toothache committed 5 days ago

Verified d46218a8

Java: release provider option values with the correct strings (#31828)

Nikhi00718 committed 5 days ago

Verified ff8d83b8

[WebGPU] Fix 1D-dispatch shader fast path (#32343)

daijh committed 5 days ago

Verified db3bc5a0

Older