Commits microsoft/onnxruntime

[CUDA] PagedAttention: quantized KV cache, XQA decode, MLA, QK-Norm and head sink (#29912)

tianleiwu committed 16 days ago

Verified e644fd56

[CUDA] Add win-arm64 packaging and size options to plugin pipeline (#31635)

tianleiwu committed 16 days ago

Verified 6df2a551

[WebGPU plugin EP] Update release and packaging-related docs (#29059)

edgchen1 committed 16 days ago

Verified c0b1d6bd

Handle empty tensors in CUDA InstanceNormalization (#31647)

apsonawane committed 16 days ago

Verified 4bae5766

[DML EP] Fix wide string handling in OpKernelInfoWrapper::GetWideName (#31656)

miaobin committed 16 days ago

Verified 4e48876e

web: fix webpack/Terser release build crash in Web CI Pipeline (#31652)

Copilot committed 16 days ago

Verified b996de6b

[CUDA] MatMulBlockQuantizedFp8Weight: fold the W8A8 activation QDQ into the decode GEMV (#31481)

tianleiwu committed 17 days ago

Verified 064861b0

[onnxruntime-web] Plans to deprecate and remove WebGL and JSEP (#29716)

edgchen1 committed 17 days ago

Verified fef99e8d

[WebNN EP] Reuse shared WASM loader for Blob-backed external data (#31151)

Honry committed 17 days ago

Verified 30c3dc47

Throw error on negative Split tensor axis [CPU] (#31149)

nenad1002 committed 17 days ago

Verified dfa30b05

[Build] update build flags of cuda plugin CI to avoid OOM (#31622)

tianleiwu committed 17 days ago

Verified 2ed68162

[WebGPU EP] Support int64 for Tile and Concat (#31049)

miaobin committed 17 days ago

Verified f6d5554e

[Build] Fix cuda plugin linux build errors (#31616)

tianleiwu committed 18 days ago

Verified 5ea5e692

[Build] Fix Windows ARM64 CUDA plugin build (#31617)

tianleiwu committed 18 days ago

Verified edb8580d

[Build] Disable SM120 TMA kernels with MSVC and CUDA 13 (#31615)

tianleiwu committed 18 days ago

Verified 4cf0af68

Fix WebGPU data transfer callbacks on Windows x86 (#31568)

mustjab committed 18 days ago

Verified 143265aa

[Build] Exclude sm120 TMA in Windows Cuda 13 plugin (#31609)

tianleiwu committed 18 days ago

Verified 5d8b70da

[CUDA] Speed up 8-bit MatMulNBits dequantization with byte permutes (#31350)

tianleiwu committed 18 days ago

Verified 5b249f65

Fix cuda 13 windows cuda plugin build (#31608)

tianleiwu committed 18 days ago

Verified 6ac1a853

[CUDA] Support bfloat16 in AllReduce, AllGather and AllToAll (#31571)

tianleiwu committed 18 days ago

Verified b2fe8b06

Fix segfault in MatMulNBits prepacking when weights/scales are parent-graph initializers in subgraph (#31141)

Copilot committed 18 days ago

Verified 03df06c1

[OpenVINO EP] Fix get/set_float_initializer_data for raw_data backed float initializers (#31138)

wangw-1991 committed 18 days ago

Verified b84e249b

[WebNN EP] Support LpNormalization op (#31197)

Honry committed 18 days ago

Verified ad094ca0

[CUDA] Coalesce QMoE MXFP4/NVFP4 weight dequantization (#31349)

tianleiwu committed 18 days ago

Verified e5d4a891

[WebGPU] Add GRU operator support (#29840)

wuisabel-gif committed 19 days ago

Verified 82a81257

[OpenVINO EP] Fix constant output mis-routing for names containing '/' (#29882)

wangw-1991 committed 19 days ago

Verified 8d187b0d

[CUDA] Block-scaled MatMul decode GEMV: tensor cores, packed FP4 decode and M-tiling (#31155)

tianleiwu committed 21 days ago

Verified 86af0f16

Fix LinearAttention output shape inference for standard GQA (q > kv) (#29892)

mirounga committed 21 days ago

Verified b725f3d3

Correct the overstated exception-free claim in ReadEpContextData (#31195)

GopalakrishnanN committed 21 days ago

Verified 7d3b4d93

[OpenVINO EP] Support dot-separated KV cache tensor names in stateful (CausalLM) transform (#29895)

zlma7001 committed 21 days ago

Verified 2aa23ccd