microsoft/onnxruntime

Pull Requests Commits

update Readme (#6903)

jywu-msft committed 4 years ago

Verified 601e04fb

Ability to fuse non-square (pruned) attention weights for BERT-like models (#6850)

mfuntowicz committed 4 years ago

Verified 9126faa3

move pipeline file and change relative path (#6882)

RandySheriffH committed 4 years ago

Verified f986ffcb

No such file or directory with --use_external_data_form and int8 (#6867)

rdzotz committed 4 years ago

Verified 107c9672

Configure session thread pool spinning preference (#6895)

RandySheriffH committed 4 years ago

Verified 679718b1

Save output tensors in bert_test_data tool (#6872)

tianleiwu committed 4 years ago

Verified 8f1786d5

Fix app packaging in UWP (#6804)

tiagoshibata committed 4 years ago

Verified fa8d1b44

Revert Gather Grad optimization in PR 6381 targeted for Rocm (#6880)

Suffian Khan committed 4 years ago

Verified 7915b670

Add check that the first 2 Loop subgraph inputs have an shape (could be explicit or inferred) as we need to know the rank the subgraph expects. Other inputs to the subgraph are more opaque so we can just pass them through. (#6891)

skottmckay committed 4 years ago

Verified 54cdb6af

Move constants from heap to stack to avoid randomness on cudnn function (#6869)

RandySheriffH committed 4 years ago

Verified d01006fc

Workaround for HTTP Error 403: Forbidden for MNIST dataset (#6885)

baijumeswani committed 4 years ago

Verified ed1883a9

[NNAPI EP] Add per-tensor u8s8 support for Qlinear[Conv/MatMul] (#6818)

guoyu-wang committed 4 years ago

Verified fedb6842

[CoreML EP] Add [Average/Max]Pool support (#6870)

guoyu-wang committed 4 years ago

Verified 3c5d811e

Support optional inputs/outputs in custom op development (#6727)

hariharans29 committed 4 years ago

Verified 9a9e741a

Add comment (#6860)

jingyanwangms committed 4 years ago

Verified f22f04a1

Reroute quantization tool readme to /docs page (#6854)

faxu committed 4 years ago

Verified 6285ee23

support opset13 in embednorm (#6866)

wangyems committed 4 years ago

Verified 9073f7a5

Change OpKernel class to be shared with shared providers (#6837)

RyanUnderhill committed 4 years ago

Verified 0d0eb2c8

Refine force CPU fallback logic in the CUDA EP (#6849)

hariharans29 committed 4 years ago

Verified 38796ad4

Add support for op kernel type control required types, require int64 for some ops (#6832)

edgchen1 committed 4 years ago

Verified 66df167a

Only report Android Baseline binary size for master branch (#6844)

guoyu-wang committed 4 years ago

Verified 36a44d55

[CoreML EP] Add Concat support (#6834)

guoyu-wang committed 4 years ago

Verified 5cf66069

Update Attention and QAttention to support pruned model (#6819)

tianleiwu committed 4 years ago

Verified 2d6e10ba

Do not create compute stream when external CUDA allocator is used. (#6833)

codemzs committed 4 years ago

Verified cb8d8464

wangyems committed 4 years ago

Verified b4b87ac7

Improvements to quantizer: Removed unused qType field, add reshape op (#6179)

pranav-prakash committed 4 years ago

Verified d5175795

Fixed issue in python cmake to update wheel package (#6384)

suryasidd committed 4 years ago

Verified 34261087

Make keepdims to its default value when adding ReduceMin/ReduceMax for quantization calibration (#6788)

chilo-ms committed 4 years ago

Verified 9b3171e9

Setup perf in docker and add features (#6582)

oliviajain committed 4 years ago

Verified db05d53b

fix issues caused by quantize/calibrate changes (#6802)

stevenlix committed 4 years ago

Verified d5f292ab

Older