Commits microsoft/onnxruntime

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

tianleiwu committed 6 days ago

Verified a06675e9

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

sushraja-msft committed 6 days ago

Verified c4f19617

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

tianleiwu committed 6 days ago

Verified e0ad0718

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

tianleiwu committed 6 days ago

Verified a1fc71e7

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

chilo-ms committed 6 days ago

Verified 5eb4aeeb

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

prathikr committed 6 days ago

Verified 7a123714

Fix rotary embedding oob issue (#29014)

apsonawane committed 6 days ago

Verified a65ac449

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

apsonawane committed 6 days ago

Verified 4819d19e

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

apsonawane committed 6 days ago

Verified fee68589

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

tianleiwu committed 6 days ago

Verified f2559a5d

Model package: fold external_data into session options and remove legacy directory load path (#29501)

jambayk committed 6 days ago

Verified eca15c62

Fix packed-QKV and broadcast-head bias strides in quantized GQA flash attention (#28963)

tianleiwu committed 6 days ago

Verified 582fe4cb

Release outputHandlesArr in OrtTrainingSession.evalStep JNI binding (#29576)

Sammy-Dabbas committed 6 days ago

Verified 5ae3496c

webgpu: Fix GQA decode split-reduce head_size out-of-bounds race (#29593)

qjia7 committed 6 days ago

Verified ced4d830

[CUDA] Fix fpA_intB internal test build: typed nullptr for TryMatMulNBits bias (#29596)

tianleiwu committed 6 days ago

Verified 233d95d0

Recover Conv/ConvTranspose rank from weight when input shape is unknown (#29149)

fanchenkong1 committed 6 days ago

Verified d5f022a3

Fix extended minimal builds with the Core ML EP (#29387)

hnsyprst committed 6 days ago

Verified 8a6eec2f

Fix wrong ORT version in Windows DLL VersionInfo for pipeline builds (#29545)

Sammy-Dabbas committed 6 days ago

Verified 1b30a664

WebGPU: enable INT64 for Equal/Sub/Where/ReduceSum under enable_int64 flag to support Gemma4 (#29392)

feich-ms committed 6 days ago

Verified ee91e35d

Avoid optional requests import in TF converter (#28825)

GopalakrishnanN committed 7 days ago

Verified 04cc7472

[Build] Fix cuda packaging build error (#29580)

tianleiwu committed 7 days ago

Verified 2125ec43

[CUDA] Support prepacked fpA_intB weights in MatMulNBits (#29499)

tianleiwu committed 7 days ago

Verified 5edd3123

[QNN-EP] add gather handler for transpose optimizer (#28755)

quic-muchhsu committed 7 days ago

Verified aeab3834

Add CudaQuantizer to onnxruntime.quantization (#29509)

tianleiwu committed 7 days ago

Verified 1990bd5f

Don't echo command when setting VSO variable in mac-cpu-packing-jobs.yml. (#29575)

edgchen1 committed 7 days ago

Verified a491809a

Guard MlasConvPrepare working-buffer products with SafeInt (#29444)

apsonawane committed 7 days ago

Verified fccf03ba

Fix a regression in graph-capture session initialization that rejects an empty graph (#29457)

adrastogi committed 7 days ago

Verified 0a62d6a1

Fix Flash and Lean attention split heuristics (#29554)

tianleiwu committed 7 days ago

Verified 4942b12b

[CUDA] Promote CUDA plugin EP as default provider (#29544)

tianleiwu committed 7 days ago

Verified 22ae7964

Honor model_external_initializers_file_folder_path for file-path model loads (#29459)

jambayk committed 7 days ago

Verified a0d9008f