microsoft/onnxruntime

Pull Requests Commits

[WebGPU] Cover mixed-scale QMoE SIMT expert offsets

aciddelgado committed 7 hours ago

ef254e5c

Change scale to use f16 type in matmul_nbits_small_m_simt

aciddelgado committed 2 days ago

Verified f29e8e8c

Change ShaderUsage for zero_points and bias inputs

aciddelgado committed 2 days ago

Verified 8a39904e

Enable default small-M INT4 SIMT for WebGPU MatMulNBits

Copilot committed 2 days ago

e03d4005

Bump electron from 42.11.9 to 42.11.16 in /js/web (#33609)

dependabot[bot] committed 2 days ago

Verified 11fdbcdc

Bump next from 15.5.27 to 15.5.28 in /js/web/test/e2e/exports/testcases/nextjs-default (#33553)

dependabot[bot] committed 2 days ago

Verified ec2db718

[CUDA] Add fast INT3 QMoE kernels and offline SM80 INT2/INT4 prepack support (#33569)

tianleiwu committed 3 days ago

Verified bd6010c0

Add exact normalization and rotary CUDA operator modes (#33469)

apsonawane committed 4 days ago

Verified c7873286

[CUDA] Fix Windows nvcc error for FP4 rescale device variable template (#33554)

tianleiwu committed 5 days ago

Verified 36f07f76

Fix QMoE SM80 FP4 init order: don't let the tactic seed throw (#32194)

GopalakrishnanN committed 5 days ago

Verified 5d466ba8

Include originating function in C-API catch-all exception message (#29561)

GopalakrishnanN committed 5 days ago

Verified 73b0fc6c

[CUDA] Add HyperConnectionLowRankMix fused low-rank hyper-connection op (#33384)

tianleiwu committed 5 days ago

Verified 245f59b7

Increase macOS CI build-and-test timeout to 150 minutes (#33564)

eserscor committed 5 days ago

Verified db2039d0

Fix combined packaging stage (#33561)

eserscor committed 5 days ago

Verified 7ec27079

[Python] Add an experimental CUDA 13 unified-wheel packaging pipeline (#33034)

tianleiwu committed 5 days ago

Verified 7c914a2a

Bump markdown-it from 14.3.2 to 14.3.3 in /js/common (#33506)

dependabot[bot] committed 5 days ago

Verified 871ad998

Bump actions/cache from 5 to 6 (#33542)

dependabot[bot] committed 5 days ago

Verified 5b8db9ab

Bump the codeql group with 4 updates (#33541)

dependabot[bot] committed 5 days ago

Verified 84829e76

Bump electron from 42.11.3 to 42.11.9 in /js/web (#33552)

dependabot[bot] committed 5 days ago

Verified f838a306

[WebGPU] Capture and replay explicit buffer clears (#33509)

qjia7 committed 5 days ago

Verified cb2a4f58

[WebGPU] Batch input uploads for built-in and plugin EPs (#33491)

qjia7 committed 5 days ago

Verified 07ac1719

Add onnxruntime_DISABLE_STREAM build option (#33482)

rvandermeulen committed 6 days ago

Verified 96dd9658

Refine WebGPU EP release process documentation. (#33520)

edgchen1 committed 6 days ago

Verified 6188bc09

Make CPU fallback type policy compile-time (#29647)

GopalakrishnanN committed 6 days ago

Verified 4e9488c3

Validate Attention projection and cache dimensions (#32946)

GopalakrishnanN committed 6 days ago

Verified 2e912580

[CUDA] Make MatMulNBits and packed attention L1 estimates route-complete (#33313)

chilo-ms committed 6 days ago

Verified 0cbe5756

P0/P1: packed-code foundation and INT3 QMoE v1 contract (#33394)

jiafatom committed 6 days ago

Verified d9c279b7

fix: scope tpause_spin_delay_cycles to the _tpause branch (#33532)

ninglinLiu committed 6 days ago

Verified fe7700a7

Group CodeQL action updates in Dependabot (#33533)

eserscor committed 6 days ago

Verified 5cd8dbb1

Fix WebGPU bundled Dawn platform cleanup (#33526)

edgchen1 committed 6 days ago

Verified 1f4b965e

Older