microsoft/onnxruntime

Pull Requests Commits

Polish Lamb Kernel (#7299)

weixingzhang committed 4 years ago

Verified c22963c2

Improve logged message for nodes that are forced to execute on CPU rather than some other EP (usually CUDA) (#7297)

hariharans29 committed 4 years ago

Verified 711cc99f

Polish Adam kernel (#7294)

weixingzhang committed 4 years ago

Verified 8ad5007f

change half gemm to use compute_32f as default (#7253)

tianleiwu committed 4 years ago

Verified 274e2fea

Support transpose by merge Reshape etc into direct xint8 operators. (#7265)

zhanghuanrong committed 4 years ago

Verified a4fdb4db

Narrow profiling scope (#7281)

RandySheriffH committed 4 years ago

Verified 42051c91

Enable CoreML EP for minimal extended mode (#7266)

guoyu-wang committed 4 years ago

Verified 370f9b88

Add support to dynamic positional/keyword input for ORTModule (#7189)

Thiago Crepaldi committed 4 years ago

Verified 7b4362c2

Fix codeql java warning (#7280)

guoyu-wang committed 4 years ago

Verified 4969431e

[Symbolic shape infer] fix scalar shape in Expand (#7285)

KeDengMS committed 4 years ago

Verified 0d49e539

NCHWc: avoid buffer reordering around Add nodes (#7279)

tracysh committed 4 years ago

Verified bc6ef809

Enable symbolic shape inference in ORTModule (#7282)

ytaous committed 4 years ago

Verified e14b291c

Suppress tracer warnings from onnx export in ORTModule (#7221)

baijumeswani committed 4 years ago

Verified d272c843

[OpenVINO-EP] Enabling save/Load blob feature (#7054)

MaajidKhan committed 4 years ago

Verified 27e77890

Add QGEMM benchmark (#7268)

chenfucn committed 4 years ago

Verified def4cc09

Restrict ConvGrad to __CUDA_ARCH__>=700 (#7278)

SherlockNoMad committed 4 years ago

Verified aa2c4651

ConvGrad CUDA Kernel Bugfix (#7273)

Lafi7e committed 4 years ago

Verified beb299e1

Support eval mode and torch.no_grad context in ORTModule and restructure ortmodule.py (#7162)

baijumeswani committed 4 years ago

Verified 844361bc

fix for using tensorrt:20.12 base image (#7264)

jywu-msft committed 4 years ago

Verified 025abf99

CUDA ConvGrad Kernel (#7227)

SherlockNoMad committed 4 years ago

Verified 4bc17ca0

Fix initializer counts when used as graph output (#7260)

TomWildenhain-Microsoft committed 4 years ago

Verified 8219518a

Remove ROCM workaround for half-to-double cast.

jessebenson committed 4 years ago

2ec452cd

Avoid passing zero bias to Gemm in gradients (#7244)

Derek Murray committed 4 years ago

Verified 25e261f1

build ONNXRuntime into WebAssembly (#6478)

fs-eire committed 4 years ago

Verified 405ca490

Not-where fusion (#7182)

ashbhandare committed 4 years ago

Verified 2aa89989

QDQ: type conversion and more ops support (#7243)

yufenglee committed 4 years ago

Verified 790fc11e

Allocate external Rocm allocator via PyBind (#7148)

raviskolli committed 4 years ago

Verified 5d759e18

Update opset for other training graphs to 12. (#7259)

Derek Murray committed 4 years ago

Verified 6308e709

Add function body to GeluGrad schema (#7190)

gramalingam committed 4 years ago

Verified a9ff4c29

Add mlas_bench tools. Starting with sconv bench and sgemm bench. (#7139)

zhanghuanrong committed 4 years ago

Verified dbcfc4be

Newer Older