microsoft/onnxruntime

Pull Requests Commits

implement split kv attention and multihead attention

aciddelgado committed 2 years ago

5fb30743

aciddelgado committed 2 years ago

cdc65dcc

[CUDA] Fix performance bug in DecoderMaskedMultiheadAttention for BeamSearch (#17613)

hariharans29 committed 2 years ago

Verified c65e8920

aciddelgado committed 2 years ago

3bf777c5

Bump Up Version to 1.17.0 (#17587)

Lafi7e committed 2 years ago

Verified e6301eee

Remove onnxruntime extensions from list of gitmodules (#17615)

nums11 committed 2 years ago

Verified f297d4df

aciddelgado committed 2 years ago

5b076f70

Update npm-packaging-pipeline.yml to always use artifacts from main branch (#17604)

fs-eire committed 2 years ago

Verified d522cc7c

[JS/WebGPU] Support If on WebGPU (#17478)

hariharans29 committed 2 years ago

Verified 460f17fb

fix illegal access memory issue

aciddelgado committed 2 years ago

2d0b960b

Avoid `get_logger` overriding root logger level (#17569)

BowenBao committed 2 years ago

Verified 152e61da

Refactor Attention cuda kernel (#17578)

tianleiwu committed 2 years ago

Verified 730fab30

Pin beartype version (#17599)

wschin committed 2 years ago

Verified 068300d9

Remove reference to internals in torch.onnx in test (#17550)

justinchuby committed 2 years ago

Verified d350ab31

Provide kwargs to remove_shared_initializers (#17539)

jambayk committed 2 years ago

Verified f969e7f8

Improve Win QNNEP pipeline (#17586)

mszhanyi committed 2 years ago

Verified 7116e66c

[js/web] FP16 binary and unary ops (#17515)

dakenf committed 2 years ago

Verified 0f406ca1

Merge remote-tracking branch 'origin/main' into aciddelgado/group_query_attention

aciddelgado committed 2 years ago

d78f4769

aciddelgado committed 2 years ago

2941dbc7

aciddelgado committed 2 years ago

3a11592a

aciddelgado committed 2 years ago

04801df4

[QNN/CPU EP] Add 16-bit Quantize/Dequantize contrib ops (#17015)

adrianlizarraga committed 2 years ago

Verified dea425e7

[ROCm] Update whisper benchmark script (#17391)

PeixuanZuo committed 2 years ago

Verified af14ae80

[TRT EP] Fix ProviderOptions functions (#17567)

simonjub committed 2 years ago

Verified c9692373

[TensorRT EP] Fallback to CUDA EP if it's explicitly assigned (#17535)

yf711 committed 2 years ago

Verified 705f8a37

[js/web] update test to explicitly fail for webnn without proxy (#17554)

fs-eire committed 2 years ago

Verified efd416b7

[js/web] update npm test to load test cases only for required backends (#17555)

fs-eire committed 2 years ago

Verified 15588759

Remove redundant Resolve() after each inlined function (#17556)

yuslepukhin committed 2 years ago

Verified fdb13264

Refactoring of attention cuda kernel: move prepare qkv and concat_past_to_present (#17559)

tianleiwu committed 2 years ago

Verified adb0be45

Update optimize_pipeline for SDXL (#17536)

tianleiwu committed 2 years ago

Verified af80542e

Older