strisunshinewentingwang/pytorch

Pull Requests Commits

Setting automatic default selection for ONNX IR v4 semantics in ONNX export API (#26146)

Spandan Tiwari committed 4 years ago

af3b15b7

Add traces to specialize_autograd and lower_grad_of (2nd try)

Krovatkin committed 4 years ago

8b126022

Expands TestAutogradDeviceType (#26708)

Mike Ruberry committed 4 years ago

a172fbf9

Remove duplicate calculation of output shape (#26684)

dskhudia committed 4 years ago

fa7b621a

Use noop observer to pass dtype for dynamic quantization (#26709)

Dmytro Dzhulgakov committed 4 years ago

128a65e2

Speed up an integer to the power of a positive integer on CPU (#26020)

xuhdev committed 4 years ago

ae0732cd

allow building docker without torchvision (#26168)

David Pollack committed 4 years ago

66d27504

Add tests for C++ functional cosine_similarity and pairwise_distance, and clean up functional test code (#26559)

Will Feng committed 4 years ago

3cae3021

Updating submodules

svcscm committed 4 years ago

714b05e4

Don't generate named tensor functions to RegistrationFunctions.h (#26685)

zou3519 committed 4 years ago

ff78d743

Typo fix (#26417)

ezyang committed 4 years ago

05f70818

resolve ignored module method type annotations (#26683)

Elias Ellison committed 4 years ago

efaa65dd

remove tools/setup_helpers/cudnn.py (#25876)

xuhdev committed 4 years ago

5e5cbcee

Add warning to anomaly_mode doc fix #26408 (#26615)

albanD committed 4 years ago

9f3351de

Vectorize unary operator erfinv (#26629)

pearu committed 4 years ago

cf1dbc79

Add derivative for cholesky_inverse (#26451)

vishwakftw committed 4 years ago

c6432909

Move the CUDA implementation of trunc to ATen. (#25423)

xuhdev committed 4 years ago

7bdc0c13

Automatic update of fbcode/onnx to 23bb6ea1a71f08e200114a153f48bd7adb66d486 (#26441)

houseroad committed 4 years ago

d6ee5849

C++ API parity: at::Tensor::set_data

pbelevich committed 4 years ago

450504cd

Use optimized graph in Inline (essentially, making Inline recursive now). (#26489)

Mikhail Zolotukhin committed 4 years ago

2cf1183e

Add 'optimized_graph' to Function. (#26488)

Mikhail Zolotukhin committed 4 years ago

c522b635

Use std::mutex instead of std::call_once in Function when we initialize GraphExecutor. (#26571)

Mikhail Zolotukhin committed 4 years ago

c034f979

Remove 'recurse' parameter from Inline. (#26487)

Mikhail Zolotukhin committed 4 years ago

76e2ffc8

Enable registering stackbased kernels with lambdas (#26658)

smessmer committed 4 years ago

a65db650

Revert D17495679: [pytorch][PR] A few hub improvements

Karl Ostmo committed 4 years ago

839e636f

Updates and extends TestNNDeviceType (#26638)

Mike Ruberry committed 4 years ago

98bbb778

Allow per-channel QTensor accept any floating type for scales (#26676)

Dmytro Dzhulgakov committed 4 years ago

ade60f8a

Per-channel quantized tensor to have only a single axis (#26675)

Dmytro Dzhulgakov committed 4 years ago

b93823cb

Fix _empty_per_channel_affine_quantized to be less hacky (#26243)

Dmytro Dzhulgakov committed 4 years ago

9aad4d7b

adding OpProfile proto into ProfDAGProtos to support storing operation cost (#26677)

wenqicaofb committed 4 years ago

fbc3c148

Newer Older