Pull Requests microsoft/onnxruntime

[CUDA Plugin EP] Expose kernel sync stream for scratch allocation

#29244 by tianleiwu was merged 2026-06-26 21:52

Fix optional-output guard in DecoderAttention/MultiHeadAttention shape inference

#29243 by titaiwangms was closed 2026-06-25 22:37

Validate SparseAttention CSR indices and key lengths element values

#29242 by yuslepukhin was merged 2026-06-26 21:56

[Pipeline] Fix CUDA Python packaging pipeline

#29241 by hariharans29 was merged 2026-06-24 00:12

Clamp derived sequence lengths and KV-cache index in CUDA GroupQueryAttention

#29240 by jiafatom was merged 2026-06-24 22:08

Fix bounds in WhisperDecoderSubgraph::CreateInitialFeeds initial feeds

#29239 by jiafatom was merged 2026-06-26 22:23

Normalize bool tensor raw_data to {0, 1} on unpack

#29238 by jiafatom was merged 2026-06-27 04:54

Fix parsing issue with WebGPU testing skill

#29237 by adrastogi was merged 2026-06-23 23:15

WebGPU: Add indirect dispatch for flash attention graph capture to support Gemma4 ep:WebGPU

#29236 by feich-ms was merged 2026-07-02 21:53

[CUDA] Support user compute stream with CUDA graph in CUDA plugin EP

#29221 by tianleiwu was merged 2026-06-25 00:20

[QNN EP] Fix segfault in PadNodeGroupSelector when dq_nodes is empty (#28717)

#29220 by yuslepukhin was closed 2026-06-23 16:52

Fix incorrect string tensor creation in test

#29219 by yuslepukhin was merged 2026-06-23 16:53

Fix GroupQueryAttention right-padded rotary prefill CUDA test

#29218 by tianleiwu was merged 2026-06-23 00:10

Revert "[CUDA] Enable XQA by default for FP16/BF16 GQA (#29046)"

#29217 by hariharans29 was closed 2026-06-22 21:34

[CPU] Add FP32 GEMV decode kernel for GroupQueryAttention

#29216 by tianleiwu was merged 2026-06-26 06:29

WebGPU: make WebGpuContextFactory::Cleanup exception-safe and idempotent

#29207 by samuel100 was closed 2026-06-26 21:17

Fix CUDA/cuDNN DLL preload paths for CUDA 13 consolidated wheel layout

#29202 by tianleiwu was merged 2026-06-26 22:52

Fix DropQDQ not dropping Q/DQ around opset-22 MaxPool

#29201 by tairenpiao was merged 2026-06-23 07:36

[Pipeline] Fix Python packaging pipeline failures

#29200 by hariharans29 was merged 2026-06-23 02:18

Fix SimplifiedLayerNorm fusion with node-produced Pow exponent

#29196 by the0cp was merged 2026-06-29 17:33

Fix TransposeOptimizer type error on zero-point-less DequantizeLinear (#28716)

#29192 by tairenpiao was merged 2026-06-20 23:16

[CUDA] Enable CUDA GQA QK-Norm and XQA decode

#29186 by tianleiwu was merged 2026-06-23 02:34

[CUDA] Add sliding-window support to XQA decode

#29177 by tianleiwu was merged 2026-06-23 21:21

[CUDA] update doc for cuda contrib op GroupQueryAttention

#29173 by tianleiwu was merged 2026-06-22 16:53

[CUDA] Fuse MoE router bias into MatMulNBits GEMV

#29170 by tianleiwu was merged 2026-06-24 09:07

[CUDA]: Split-K2 QMoE SwiGLU GEMV kernel

#29167 by tianleiwu was merged 2026-06-30 05:44

[CUDA] Default QMoE GEMV fp16 accumulation for fp16 activations

#29166 by tianleiwu was merged 2026-06-19 22:48

[CPU] Enable pre-packed weights sharing for MatMulNBits

#29163 by derdeljan-msft was merged 2026-06-26 22:25

[CUDA] Enable XQA decode for GroupQueryAttention with attention sink

#29162 by tianleiwu was merged 2026-06-20 00:32

[CUDA] Optimize FlashDecode split planning for local-window GQA

#29161 by tianleiwu was merged 2026-06-19 18:16