microsoft/onnxruntime

Pull Requests Commits

rotary dim 1 fix

aciddelgado committed 1 year ago

0a18189c

Merge branch 'main' into aciddelgado/fix_rotary_gqa

aciddelgado committed 1 year ago

806e0f02

[EP Perf] Add tag to indicate which TRT parser is using (#19784)

yf711 committed 1 year ago

Verified 3170a48e

Update script to run CIs for a branch. (#19797)

skottmckay committed 1 year ago

Verified 01c376a0

[JS/WebGPU] Preserve zero size input tensor dims. (#19737)

satyajandhyala committed 1 year ago

Verified 24b72d26

Run CoreML EP with NeuralNetwork and ML Program in CI unit tests (#19796)

skottmckay committed 1 year ago

Verified 6c3bed67

Implement IsNaN-9,13,20 for CUDA along with tests (#19807)

yuslepukhin committed 1 year ago

Verified 29643526

Remove memset for the case no any mask (#19823)

yihonglyu committed 1 year ago

Verified 33578cc7

Fix argparser in `matmul_bnb4_quantizer` (#19812)

jambayk committed 1 year ago

Verified 3dfce2f1

cuda graph enhancement (#19636)

wangyems committed 1 year ago

Verified 72ce4de0

Update tolerance of provider tests to fix flaky tests (#19792)

tianleiwu committed 1 year ago

Verified bff4f8bf

Define recomputable op list with domain/opset (#19722)

pengwa committed 1 year ago

Verified 5c5d6e99

[WebNN EP] Make sure optional input is provided (#19686)

Honry committed 1 year ago

Verified 1ce5bfb0

Add SpaceToDepth and DepthToSpace CUDA NHWC Ops (#19646)

mtavenrath committed 1 year ago

Verified f2dc725b

Fix GQA Rotary Embedding sequence length (#19801)

aciddelgado committed 1 year ago

Verified 8bd1335d

reset dcvsEnable for different HTP performance mode (#19728)

HectorSVC committed 1 year ago

Verified db8d0c8e

modify rotary cache dim check

aciddelgado committed 1 year ago

b43de5ef

Upgrade the Windows SDK version that is used in WindowsAI Nuget Packaging pipeline (#19786)

Changming Sun committed 1 year ago

Verified f9a92e58

Adapt memory optimizer to fit PHI2 (#19757)

pengwa committed 1 year ago

Verified d9bf8561

Remove arm build for training (#19788)

askhade committed 1 year ago

Verified e93a8608

Don't reduce warning level for CUDA build on Windows (#19663)

skottmckay committed 1 year ago

Verified db59cec8

[js/web] dump debug logs for karma for diagnose purpose (#19785)

fs-eire committed 1 year ago

Verified a7885140

ATen Op Supports Int Return Type and CPU Tensor Arguments (#19773)

Lafi7e committed 1 year ago

Verified 1bfc2668

Fix seed for recomputed Dropout (#19715)

pengwa committed 1 year ago

Verified d1025697

[TensorRT EP] Fix bug for DDS output handling for empty tensor (#19575)

chilo-ms committed 1 year ago

Verified d9730c7f

Implement CUDA IsInf-10,20 (#19772)

yuslepukhin committed 1 year ago

Verified 1e78bcea

Adding cuda kernel (optimized for sm80) for block-wise 4b quantized float 16 GEMM. (#18619)

chenfucn committed 1 year ago

Verified 06e684c9

Fix CUDA BatchNorm bugs and add support for NHWC (#19742)

mtavenrath committed 1 year ago

Verified bdf678df

enable embedding sparse optimization by default (#19714)

guyang3532 committed 1 year ago

Verified cd56ea4a

[quant] supports act_order inputs in Matmulnbits and new quantization algorithm "hqq" (#19106)

wejoncy committed 1 year ago

Verified 7e613ee8

Older