Commits microsoft/onnxruntime

Fix empty-window OOB read in CUDA MaxPool index kernel; add regression tests

Copilot committed 5 days ago

Verified ba67e47f

Clarify comments per review feedback

Copilot committed 5 days ago

Verified f944d971

Fix CUDA MaxPool dilation phase on negative window start; add parity tests

Copilot committed 5 days ago

Verified b7ac7212

Initial plan

Copilot committed 5 days ago

Verified 258d9817

[webgpu] Enable im2col-matmul for newer Xe platforms; limit to fp16 (#29505)

daijh committed 5 days ago

Verified 632de623

webgpu: Bump FlashAttentionDecodeQKV workgroup to 128, tile_size_k_vec to 32 (#29586)

qjia7 committed 6 days ago

Verified 1ad2a1f2

[ORT] onnxruntime_perf_test: Add multi-shape profiling support to perftest tool (--data_shape) (#29571)

ankitm3k committed 6 days ago

Verified 4d8886f8

Guard GridSample CUDA coordinate conversion (#29581)

GopalakrishnanN committed 6 days ago

Verified f631f0d9

Avoid throwing C++ exceptions from WebGPU EP Dawn callbacks (#29625)

edgchen1 committed 6 days ago

Verified c57e0e50

Add IOBinding test for preallocated CPU output buffer reuse (#29566)

GopalakrishnanN committed 6 days ago

Verified 1c34fd5c

Avoid small MatMul batch parameter heap allocations (#29085)

GopalakrishnanN committed 6 days ago

Verified dbe7655d

Fix libcudart.so.13 hard dependency in pybind module breaking import on CPU-only Linux (#29590)

Copilot committed 6 days ago

Verified dd32f359

Handle non-finite ROI coordinates in CropAndResize (#29605)

titaiwangms committed 6 days ago

Verified bcc9d30b

Validate DynamicQuantizeLSTM recurrence quantization parameter shapes (#29254)

titaiwangms committed 6 days ago

Verified 682936c1

Speed up CUDA build by splitting SM80 MoE GEMM generated files (#29614)

tianleiwu committed 6 days ago

Verified 5dc8d4c6

Skip MinLength logits processor construction when eos_token_id is negative (#29604)

titaiwangms committed 6 days ago

Verified 4ce84a76

[CUDA] Support attention_bias in GroupQueryAttention via the unfused path (#29525)

jiangzhuo committed 7 days ago

Verified 2cd57057

webgpu: two optimizations for the subgroup Gemm/MatMul kernels: (#29271)

xhcao committed 7 days ago

Verified 02789a58

chore: bump version to 1.29.0 (#29600)

sanaa-hamel-microsoft committed 7 days ago

Verified b0651dd5

[WebGPU] Fix OrtReleaseEnv self-deadlock after failed adapter request on Linux (#29591)

Copilot committed 7 days ago

Verified ef73996a

Fix web e2e (npm/vite) and Python DML CI pipelines (#29609)

tianleiwu committed 7 days ago

Verified a06675e9

Bump onnx to 1.22.0 and protobuf to 6.33.5 to fix security CVEs (#29606)

sushraja-msft committed 7 days ago

Verified c4f19617

[CUDA] Enable native SM90, block_size=32, and fused bias for fpA_intB MatMulNBits (#29585)

tianleiwu committed 7 days ago

Verified e0ad0718

[CUDA] Fix QMoE profiler cross-stream race and CUDA-graph-capture safety (#29584)

tianleiwu committed 7 days ago

Verified a1fc71e7

Fix CustomOp forward compatibility: cap version instead of rejecting (#29574)

chilo-ms committed 7 days ago

Verified 5eb4aeeb

adjusts conv kernel to use get/set by offset helpers (#29463)

prathikr committed 7 days ago

Verified 7a123714

Fix rotary embedding oob issue (#29014)

apsonawane committed 7 days ago

Verified a65ac449

Validate B/scales/zero_points shape in MatMulNBits::PrePack (#29445)

apsonawane committed 7 days ago

Verified 4819d19e

Fix signed-int overflow in SamplingState::Init to prevent heap-buffer-overflow (#29443)

apsonawane committed 7 days ago

Verified fee68589

Clamp 1D attention mask_index to valid bounds (#29449)

tianleiwu committed 7 days ago

Verified f2559a5d