Commits microsoft/onnxruntime

[WebGPU] Split Tan's pi/2 reduction in four parts for accuracy near poles

Aditya Rastogi committed 2 days ago

5c96e113

[WebGPU] Evaluate Tan with an exact argument reduction

Aditya Rastogi committed 2 days ago

15da50d4

[WebGPU] Evaluate Exp in f32

Aditya Rastogi committed 2 days ago

20a12e10

[WebGPU] Evaluate Erf in f32

Aditya Rastogi committed 2 days ago

8b5f85e7

[WebGPU] Keep InstanceNormalization scale and shift in f32

Aditya Rastogi committed 2 days ago

48102b53

[WebGPU] Fix f16 reduction overflow and ReduceLogSumExp overflow

Aditya Rastogi committed 2 days ago

9454f307

[CUDA] Optimize INT3 QMoE decode/prefill kernels and S=8 GEMV finalize dispatch (#33585)

tianleiwu committed 2 days ago

Verified c026a509

[CUDA] Extend M=8 narrow CtaN=2 in fpA_intB_gemv to BF16 and INT8 small-N projections (#33600)

tianleiwu committed 2 days ago

Verified f30f7fcb

[WebGPU] Use Queue.WriteBuffer for small CPU-to-GPU uploads (#33551)

aciddelgado committed 2 days ago

Verified 49481a71

[WebGPU] Batch streamless GPU-to-GPU copy vectors (#33567)

aciddelgado committed 2 days ago

Verified af0a4bab

[CUDA] Optimize GatedDeltaNetChunkedKernel with ldmatrix/SyncBeforeStore and add GatedRMSNormWarp128Kernel (#33601)

tianleiwu committed 2 days ago

Verified 1e3c937d

Update Pillow minimum version to 13.0.0 (#33656)

mustjab committed 2 days ago

Verified 7201c965

Deprecate provider-bridge legacy NvTensorRTRTXExecutionProvider (#32276)

keshavv27 committed 2 days ago

Verified 979b6d1d

[WebGPU] Optimize DP4A for large Qualcomm subgroups (#33560)

gyagp committed 2 days ago

Verified 034e0eaa

[CUDA] Include SM103 in Linux aarch64 CUDA 13 packages (#33649)

haoxiz-nvidia committed 2 days ago

Verified edb399aa

[WebGPU] Fix subgroup matrix configs not populated in WASM builds (#33531)

kylo5aby committed 2 days ago

Verified c130fb69

[CPU/WebGPU] Support empty MatMulNBits weights (#33559)

qjia7 committed 2 days ago

Verified 03a19a74

Extend sparse attention caches and optimize packed indexing (#33472)

apsonawane committed 2 days ago

Verified a23131fe

Extend GDN and causal convolution fused state handling (#33471)

apsonawane committed 2 days ago

Verified 6d63e0e0

Use image-cached Temurin JDK for CodeQL (#33653)

eserscor committed 2 days ago

Verified c3a57386

Honor explicit rotary positions in CUDA GroupQueryAttention fast decode (#32951)

BBaditi committed 2 days ago

Verified 59e37c29

[CUDA][PagedAttention] Fuse small-batch sequence-length sanitizer and add session.paged_attention.skip_input_sanitizer (#33597)

tianleiwu committed 3 days ago

Verified 55ab2dff

[CUDA][QMoE] Wire LaunchFusedRouteMaps into CutlassMoeFCRunner for S<=8 and E<=512 (#33596)

tianleiwu committed 3 days ago

Verified a3c3cf98

[CUDA][QMoE] Use PickPrefillTactic with Ampere 64x128x64 fallback for INT4 deterministic prefill (#33595)

tianleiwu committed 3 days ago

Verified cad4827e

[CUDA] Optimize HyperConnectionLowRankMix for S<=8 speculative decode and deterministic prefill (#33548)

apsonawane committed 3 days ago

Verified 81204341

[CUDA] Add SpeculativePrefixCommit contrib operator (#33474)

apsonawane committed 3 days ago

Verified 1b6e8e06

[CUDA] Add BFloat16 ArgMax/ArgMin support and extend GatedAdd with projected gate (#33470)

apsonawane committed 3 days ago

Verified 8c8e154f

Bump setuptools from 78.1.1 to 84.1.0 in /tools/ci_build/github/linux/docker/scripts (#33606)

dependabot[bot] committed 3 days ago

Verified 997d2d1b

Bump actions/checkout from 6 to 7 (#33386)

dependabot[bot] committed 3 days ago

Verified 7c6db8e6

Add executable EPContext fixture for encrypted binding tests (#33216)

GopalakrishnanN committed 3 days ago

Verified 7617fc8c