microsoft/onnxruntime

Pull Requests Commits

[JS/WebGPU] Improve MatMulNBits perf (#19974)

satyajandhyala committed 1 year ago

Verified b33216be

LLVM16 compat changes (#20294)

Changming Sun committed 1 year ago

Verified 794d39a9

Integration with ONNX 1.16.0 (#19745)

liqunfu committed 1 year ago

Verified cd7112f8

[DML EP] Move operators to feature level 6.4 (#20290)

PatriceVignola committed 1 year ago

Verified a0d50673

[QNN EP] Use QNN's ResizeBilinear operator for specific configs of ONNX Resize (#20292)

adrianlizarraga committed 1 year ago

Verified 327fb1fd

Bump protobufjs from 7.2.4 to 7.2.5 in /js/web (#20270)

dependabot[bot] committed 1 year ago

Verified 9ca1afa2

[WebNN EP] Support Gelu op (#20240)

Honry committed 1 year ago

Verified 667d2eb8

[DML] Add FastGelu (#20066)

PatriceVignola committed 1 year ago

Verified 12042a93

[js/web] support SimplifiedLayerNorm and SkipSimplifiedLayerNorm (#20277)

fs-eire committed 1 year ago

Verified 50bd4571

TRT detailed log and strong typed networks (#19695)

gedoensmax committed 1 year ago

Verified 2d0e1df8

Re-enable MatMul QDQ fusions with the DML EP (#20248)

jeffbloo committed 1 year ago

Verified f7e2faf9

[WebNN EP] Fixed WebNN constant operand is detached issue (#20229)

Honry committed 1 year ago

Verified ee603ee3

fix build break in kernel_explorer (#20235)

yufenglee committed 1 year ago

Verified e6ca3606

[TensorRT EP] Fix updating provider options from session options (#20246)

chilo-ms committed 1 year ago

Verified 47072509

[js/webgpu] implement DepthToSpace operator in webgpu (#19948)

MasayoshiTsutsui committed 1 year ago

Verified 6a9d8a90

Reduce heap contention in Tokenizer (#20196)

yuslepukhin committed 1 year ago

Verified 89a96bdc

[EP Perf] Customize onnx-tensorrt commit id when init CI tasks (#20175)

yf711 committed 1 year ago

Verified 9577fe45

Check padding density by input of embedding module (#19821)

guyang3532 committed 1 year ago

Verified 471e969e

Set parallel count to avoid OOM in training GPU packaging pipeline (#20255)

mszhanyi committed 1 year ago

Verified 0acde115

Prompt layer-wise recompute when applicable (#20126)

pengwa committed 1 year ago

Verified 280b2634

Reuse T4 for Cuda12.2 training packaging pipeline. (#20244)

mszhanyi committed 1 year ago

Verified 14d7872c

Reduce Heap contention in StringNormalizer (#20182)

yuslepukhin committed 1 year ago

Verified 7d8dea9f

Optimize constant sharing perf (#20143)

pengwa committed 1 year ago

Verified 81005e2c

Add INT16 and UINT16 compatibility for relu_quantizelinear (#20187)

danyue333 committed 1 year ago

Verified 07b5377f

Support more ops for recompute (#20234)

pengwa committed 1 year ago

Verified 41acd8c5

Fix Mixtral Parity test to keep it consistent with Transformers. (#20210)

AdamLouly committed 1 year ago

Verified 22a61a3c

fix "4bit quantization scales and zeropoint tensor shape" (#19986)

wejoncy committed 1 year ago

Verified 908a76d6

[js/webgpu] Implement com.microsoft.RotaryEmbedding (#20209)

hujiajie committed 1 year ago

Verified 23d3afd4

Improve KE for commandline and programmatically tuning dispatch (#18778)

cloudhan committed 1 year ago

Verified e19c7789

Support seq_len > 64K in rotary embedding cuda kernel (#20204)

wangyems committed 1 year ago

Verified cc3faba6

Older