Pull Requests microsoft/onnxruntime

Fix data race in CUDA graph tests: add SynchronizeInputs/SynchronizeOutputs

#29398 by Copilot was merged 2026-06-30 00:58

Handle non-trivially-copyable types in Loop/Scan output concatenation

#29397 by yuslepukhin was merged 2026-06-30 21:12

[CoreML] Address "model_path" must not be empty

#29394 by yuslepukhin was merged 2026-06-30 17:46

WebGPU: enable INT64 for Equal/Sub/Where/ReduceSum under enable_int64 flag to support Gemma4 ep:WebGPU

#29392 by feich-ms was merged 2026-07-07 04:14

cuda plugin version 0.1.0

#29390 by tianleiwu was merged 2026-06-30 00:05

[CUDA] Deprecate SkipLayerNorm strict mode

#29388 by tianleiwu was merged 2026-06-30 03:38

Fix extended minimal builds with the Core ML EP

#29387 by hnsyprst was merged 2026-07-07 07:55

Fix FP8 (FLOAT8E4M3FN) quantization scale using wrong reference distribution

#29350 by Osamaali313 was merged 2026-06-27 21:20

[CUDA] Remove unused code in moe_kernels.cu‎

#29295 by tianleiwu was merged 2026-06-27 04:55

Fix TreeEnsemble target id validation

#29293 by edgchen1 was merged 2026-06-29 22:42

fix(ci): incorrect identity for azcopy

#29274 by sanaa-hamel-microsoft was merged 2026-06-26 16:58

webgpu: two optimizations for the subgroup Gemm/MatMul kernels:

#29271 by xhcao was merged 2026-07-08 04:33

Fix optional-output guard in DecoderAttention/MultiHeadAttention shape inference

#29268 by titaiwangms was merged 2026-06-30 20:52

Fix Windows CUDA Plugin EP and CUDA CI azcopy authentication failures

#29267 by Copilot was closed 2026-06-26 15:37

Fix Windows Cuda 13.3 build for VS 2026

#29266 by tianleiwu was merged 2026-06-29 17:04

Validate initializer data length and guard element-count computation in contrib Range shape inference

#29265 by titaiwangms was merged 2026-07-09 21:53

Reject CUDA BERT EmbedLayerNorm/SkipLayerNorm shapes exceeding 32-bit output indexing

#29264 by titaiwangms was merged 2026-06-30 20:40

chore(ci): switch to github workflow dedicated docker registry for cache

#29262 by sanaa-hamel-microsoft was closed 2026-06-25 16:01

Validate GatherBlockQuantized inputs (CUDA)

#29259 by titaiwangms was closed 2026-06-25 16:45

Validate GatherBlockQuantized inputs (CUDA)

#29258 by titaiwangms was closed 2026-06-25 01:50

Validate EmbedLayerNormalization indices and use 64-bit offsets in CUDA BERT LayerNorm/SkipLayerNorm

#29257 by titaiwangms was closed 2026-06-25 16:45

Validate EmbedLayerNormalization embedding indices (CUDA)

#29256 by titaiwangms was closed 2026-06-25 01:46

Validate DynamicQuantizeLSTM recurrence quantization parameter shapes

#29254 by titaiwangms was merged 2026-07-08 17:14

Validate MaxpoolWithMask kernel rank matches input spatial rank

#29253 by titaiwangms was merged 2026-06-30 16:46

[CUDA] Make cuDNN optional at runtime

#29252 by tianleiwu was merged 2026-06-30 23:59

Fix unbounded lifetime on WithOutputTensor in Rust bindings

#29251 by sayanshaw24 was merged 2026-06-26 16:46

Validate and inline external data in node tensor attributes during session initialization

#29250 by yuslepukhin was merged 2026-06-30 21:20

Fix integer overflow in RKNPU implicit bias allocation

#29249 by GopalakrishnanN was merged 2026-06-26 22:43

[VitisAI] Recognize native-endian in-memory address tag in process_ext_address

#29248 by mingyueliuh was merged 2026-06-30 23:45

webgpu: Enable FlashAttention for batched GQA with right-padded prompts

#29247 by qjia7 was merged 2026-06-26 05:32