microsoft/onnxruntime

Pull Requests Commits

Fusion required for layoutxlm model. This model is based on BERT model but has an extra Add after the QK Matmul but the functionality is different than the existing one.

apsonawane committed 1 year ago

7d8dd541

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

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

Zhijxu/fix conv1d replacement (#19758)

zhijxu-MS committed 1 year ago

Verified 2a5c9b86

Expose SessionOtions.DisablePerSessionThreads (#19730)

yuslepukhin committed 1 year ago

Verified 0cdf36fa

[DML] MatrixMultiplyIntegerToFloat (#19608)

AnaghaRaoAMD committed 1 year ago

Verified 27b1dc91

fix split shape inference error for opset >= 13 (#19756)

inisis committed 1 year ago

Verified 2e13d5f0

Benchmark - Updating llama-2 requirement files (#19716)

ironmangithub committed 1 year ago

Verified 9acaf534

Update copying API header files (#19736)

mszhanyi committed 1 year ago

Verified 9460597b

[QNN Quant] Add preprocessing option to transpose graph inputs/outputs to channel-last (#19731)

adrianlizarraga committed 1 year ago

Verified 2d79052e

[WebNN EP] Add contraints for MatMul (#19713)

zesongw committed 1 year ago

Verified de3158e7

Enable CPUINFO for all Windows build (#19655)

Changming Sun committed 1 year ago

Verified a0521f89

[js/web] transfer input buffer back to caller thread (#19677)

fs-eire committed 1 year ago

Verified f06164ef

disable gemm f16 on CPU (#19744)

yufenglee committed 1 year ago

Verified 22176a5f

Update google benchmark to 1.8.3. (#19734)

edgchen1 committed 1 year ago

Verified 5672cdeb

Change webgpu CI pipeline to use a preinstalled chrome (#19729)

Changming Sun committed 1 year ago

Verified ed550b5f

Older