strisunshinewentingwang/pytorch

Pull Requests Commits

[quant] Disable qnnpack test when TSAN is enabled (#38153)

supriyar committed 4 years ago

13847638

Fix overflow in torch.remainder when dividend is very large (#37758)

xwang233 committed 4 years ago

63b1ae69

s/callUnboxed/call/ (#37999)

Basil Hosmer committed 4 years ago

fdc40616

Add test jobs on CPU agents for CUDA builds on Windows (#37904)

peterjc123 committed 4 years ago

55de7c3b

[JIT]Support List variable in adv indexing. (#37966)

Ailing Zhang committed 4 years ago

e84aa021

Vectorize non-persistent Softmax kernels (#36485)

slayton58 committed 4 years ago

c879c6fb

Migrate OwnerRRef value store to generic torch Future (#38143)

xush6528 committed 4 years ago

615235fc

Add documentation for FeatureAlphaDropout (#36295)

Donna Choi committed 4 years ago

ca2206d0

Fix a minor typo in DistanceOpsKernel.cpp (#37596)

yoshnary committed 4 years ago

c13dc2ca

[TensorExpr] Fix a bug in the IR Simplifier that could introduce a division by zero (#38055)

nickgg committed 4 years ago

ad433e20

int8 fc with tests (#38017)

Hector Yuen committed 4 years ago

9957db22

Add documentation for nn.Hardsigmoid and nn.functional.hardsigmoid. (#38120)

zou3519 committed 4 years ago

172bcdb8

Dont store redundant packed params in dynamic quantized RNN (#38134)

James Reed committed 4 years ago

41572116

[TensorExpr] Don't include aten::rand_like to TE fusion groups since we can't handle rand+broadcast case yet. (#38132)

Mikhail Zolotukhin committed 4 years ago

4784af1d

fix compilation error with gcc 5.5 (#38112)

cloudhan committed 4 years ago

6e1e2a60

`unfold_backward` gets its own kernel (#36612)

nikitaved committed 4 years ago

a7c29dbf

[quant][graphmode][refactor] Split quantization.cpp (#37975)

jerryzh168 committed 4 years ago

0ed7fc58

[quant][graphmode][refactor] Remove unused code in quantization.cpp (#37974)

jerryzh168 committed 4 years ago

ff9a809c

Back out "Revert D20229168: [quantization] Use torchbind for Linear PackedParams" (#38101)

James Reed committed 4 years ago

c1e7758b

[TensorPipe] Do not require user to provide worker name-to-rank map (#38052)

lw committed 4 years ago

91f451a5

Don't use Profiler key in lite interpreter (#37962)

Ilia Cherniavskii committed 4 years ago

b4946b96

Replace torch.is_tensor usages with isinstance checks. (#38062)

rgommers committed 4 years ago

726aa713

remove uses of type() and type_as() part 1. (#38029)

Ailing Zhang committed 4 years ago

9232356e

Revert D21449612: [pytorch][PR] Migrate AT_DISPATCH_ALL_TYPES_AND_COMPLEX_AND3 to c10::complex

Mike Ruberry committed 4 years ago

0c936f94

[TensorExpr] Correctly print 'bool' dtype in Cuda printer. (#38077)

Mikhail Zolotukhin committed 4 years ago

0f60c8d8

[TensorExpr] Don't include prim::Constant nodes with Tensor type into TE fusion groups - we can't handle them. (#38105)

Mikhail Zolotukhin committed 4 years ago

ff1a627b

[TensorExpr] Properly handle Bool dtype in several other places. (#38104)

Mikhail Zolotukhin committed 4 years ago

a253ea92

[TensorExpr] Correctly print dtypes in Cast and Allocate. (#38091)

Mikhail Zolotukhin committed 4 years ago

459f14e9

[tensorboard] Let hparam render values correctly (#31544)

lanpa committed 4 years ago

609d5a44

Run QEMU to test that default dispatch doesn't use AVX (#38094)

malfet committed 4 years ago

4c358b8b

Newer Older