microsoft/onnxruntime

Pull Requests Commits

Fall back from cuDNN SDPA in GQA when the cuDNN planner rejects the graph

Copilot committed 4 days ago

Verified 397bbbcc

Handle out-of-range antialiased Resize coordinates (#33056)

apsonawane committed 5 days ago

Verified ba9e4b10

Validate Attention prepack dimensions (#33061)

apsonawane committed 5 days ago

Verified 6eac349a

Support standalone Dawn API packages for native WebGPU hosts (#33006)

mustjab committed 5 days ago

Verified c0ef92bf

[MLAS] Fix fp16 overflow/NaN bugs in ARM64 SVE Gelu/Erf and add test coverage (#33040)

vijay-kapse committed 5 days ago

Verified 62ac19ab

[CUDA] Address cuDNN paged SDPA review feedback (#32624)

tianleiwu committed 5 days ago

Verified e721a97e

Validate DML upload heap allocations (#32745)

apsonawane committed 5 days ago

Verified dff72104

[MLAS] Exercise ARM64 SymmQgemm Lit (Ld64) and Big kernels in unit tests (#33041)

yairbenavraham committed 5 days ago

Verified e2d98ddb

Bump esbuild and vite in /js/web/test/e2e (#33012)

dependabot[bot] committed 5 days ago

Verified 999e887a

Bump github/codeql-action/upload-sarif from 3.37.7 to 4.38.2 (#33025)

dependabot[bot] committed 5 days ago

Verified f05448ff

Reject recursive function graph attributes (#32864)

danielsongmicrosoft committed 5 days ago

Verified 27f3d47e

Validate inline raw initializer size before unpacking (#32866)

danielsongmicrosoft committed 5 days ago

Verified 5f052c79

[CPU] Support independent scale and output types in DequantizeLinear (#32728)

TANGBUDU committed 5 days ago

Verified 43e2046f

[CUDA] Fix QMoE MXFP4 W4A16 on SM120/SM121 and speed up FP4 decode (#33067)

tianleiwu committed 5 days ago

Verified 3886abd4

[WebGPU] Add per-head Q/K normalization to PagedAttention (#32808)

aciddelgado committed 5 days ago

Verified c971ca26

[CUDA] Enable SM80+ FP16/BF16 mixed INT2/INT4 QMoE prefill by default (#33005)

jiafatom committed 5 days ago

Verified 2cdff3ca

Run targeted MLAS tests across CPU architectures (#33054)

Copilot committed 5 days ago

Verified e1b89964

[CUDA] Add SparsePagedAttention (#32524)

Copilot committed 5 days ago

Verified d32356b3

[CUDA] Avoid repacking NVFP4 QMoE decode weights (#32325)

tianleiwu committed 5 days ago

Verified f0843448

[CUDA] Fix Protobuf warning in CUDA 13.3 plugin build (#32854)

tianleiwu committed 5 days ago

Verified 7da55552

[CUDA] Implement SparseAttentionIndexer operator (#32526)

Copilot committed 5 days ago

Verified 5bbba228

Add Qwen4-Exp support to VarlenNGramHashMapping (#32467)

Copilot committed 5 days ago

Verified 38c3db5c

[CUDA] Follow up INT2 grouped GEMM test coverage and plugin build (PR4a2) (#33045)

jiafatom committed 6 days ago

Verified 78c83491

Fix packed GatedDeltaNet shape inference (#32708)

kunal-vaishnavi committed 6 days ago

Verified 12e7b2b6

Add weightless hyper-connection fusion operators (#32687)

Copilot committed 6 days ago

Verified 7001d9a1

Bump actions/setup-java from 5.7.0 to 6.0.1 (#33028)

dependabot[bot] committed 6 days ago

Verified e368af09

Handle optimizer size overflow without exceptions (#32875)

mustjab committed 6 days ago

Verified 52e8078b

[CUDA] Add DynamicSparseAttention (#32525)

Copilot committed 6 days ago

Verified 03ef37f5

Add VarlenNGramHashMapping for packed/varlen batching support in DeepSeek Engram (#32358)

Copilot committed 6 days ago

Verified e0623477

Bump brace-expansion from 1.1.16 to 1.1.21 in /onnxruntime/test/wasm (#33038)

dependabot[bot] committed 6 days ago

Verified 9ac464f6

Older