microsoft/onnxruntime

Pull Requests Commits

Enable grouped INT8 sparse attention for four-row verification

apsonawane committed 6 days ago

ebac9e56

Merge main into asonawane/main-ops-sparse-attention-indexer

apsonawane committed 6 days ago

b42576e6

Fix WebGPU bundled Dawn platform cleanup (#33526)

edgchen1 committed 6 days ago

Verified 1f4b965e

Bump electron from 41.10.6 to 42.11.3 in /js/web (#33522)

dependabot[bot] committed 6 days ago

Verified 34c34a35

Bump vite from 7.3.6 to 7.3.7 in /js/web/test/e2e (#33507)

dependabot[bot] committed 6 days ago

Verified 3d7ea7ea

Bump vite from 6.4.2 to 6.4.5 in /js/web/test/e2e/exports/testcases/vite-default (#33521)

dependabot[bot] committed 6 days ago

Verified 4aa7c7c8

Preserve distinct epsilons in GQA pre-norm fusion (#33363)

gyagp committed 7 days ago

Verified 9a43d1c9

[CUDA] Optimize dense prefill for NVFP4 hybrid QSA/GDN models (#33367)

tianleiwu committed 7 days ago

Verified e460778c

Allow OpenVINO CPU EP discovery on non-Intel CPUs (#31981)

nietras committed 7 days ago

Verified 8755f3ef

Bump follow-redirects from 1.16.0 to 1.16.1 in /js/web (#33496)

dependabot[bot] committed 7 days ago

Verified e710fb2a

Bump follow-redirects from 1.16.0 to 1.16.1 in /onnxruntime/test/wasm (#33494)

dependabot[bot] committed 7 days ago

Verified f3a2b3c0

Add compact QSA selection and specialized INT8 attention dispatch

apsonawane committed 7 days ago

d21946e6

[CUDA] Add state_status output to PackedSparseAttentionIndexer (#33383)

tianleiwu committed 7 days ago

Verified 2d8060a8

[CUDA] Honor deterministic compute for MatMulNBits tactic selection (#32898)

kunal-vaishnavi committed 7 days ago

Verified ed8f79bc

Avoid quadratic allocation-planner free-buffer scans (#33407)

rui-ren committed 7 days ago

Verified c33b4eb5

[CUDA] Cosmos3 - Fix variance cancellation in fused BERT layer normalization (#33357)

rui-ren committed 7 days ago

Verified f7b31b70

Fix CUDA LayerNorm/RMSNorm silent wrong output for rank-reduced scale (#28971)

justinchuby committed 7 days ago

Verified a926b837

Fix Gemm transpose fusion with public input outputs (#33484)

xgbtxy committed 7 days ago

Verified db5f5eaf

[WebGPU] Add bit-exact convolution state replay for speculative decoding (#33493)

aciddelgado committed 7 days ago

Verified 8b17e671

[Framework/CUDA] Add generic kernel dispatch policy reader (#33042)

chilo-ms committed 7 days ago

Verified 314ff8d4

Fix correctness guards in training graph optimizers (#32982)

Yuhx141 committed 7 days ago

Verified c55286e5

Validate ORT format edge slots (#32200)

apsonawane committed 7 days ago

Verified b923407c

[CICD Stability] Start transition to scaleset based providers (#33492)

eserscor committed 8 days ago

Verified 973f680e

[WebGPU] Automate Dawn updates (#33396)

gyagp committed 8 days ago

Verified 6a5a08c1

[WebGPU] Package D3D12 Agility SDK for Python wheels (#32795)

mei1127 committed 8 days ago

Verified fb8bfaaa

Update transformers and FlatBuffers Go example grpc dependency pins (#33501)

mustjab committed 8 days ago

Verified 53e61f3e

[CUDA] Add fused decode routing, single-pass finalize, and PDL overlap for NVFP4 QMoE (#33366)

tianleiwu committed 8 days ago

Verified 942356fc

Validate PackedAttention token offsets (#33330)

apsonawane committed 8 days ago

Verified a243f8cb

cuda pagedattention seqlens shape fix (#31651)

apsonawane committed 8 days ago

Verified 786177ab

Add opset 28 SpaceToDepth support (#32362)

justinchuby committed 8 days ago

Verified 4c184ac4

Older