strisunshinewentingwang/pytorch

Pull Requests Commits

[NNC] Make normalization transformation in-place (#56158)

navahgar committed 3 years ago

b387f7ca

[pytorch][PR] Automated submodule update: tensorpipe (#56348)

Lucas Hosseini committed 3 years ago

22d4d9f4

[CUDA graphs] Allows DeviceCachingAllocator to capture cross-stream memory use (#55860)

mcarilli committed 3 years ago

ffdecc1a

Porting logcumsumexp tests to OpInfo (#56135)

iramazanli committed 3 years ago

3e42da09

[c10d] Remove deprecated use of torch.LongTensor, torch.ByteTensor (#55861)

rohan-varma committed 3 years ago

ce05b7a3

Short circuits DistributedDataParallel._recursive_to's copy and stream syncs if input is already on the right device (#55624)

mcarilli committed 3 years ago

a24b1724

[NNC] Fuse loops that have the same bounds as expressions (#55997)

navahgar committed 3 years ago

29c5cb79

OpInfo porting for logsumexp operator (#55520)

iramazanli committed 3 years ago

b0e0841f

Vectorize copysign on CPU (#51792)

peterbell10 committed 3 years ago

8c74e1b8

Added OpInfos for eq, ne, ge, gt, le, and lt (#55709)

imaginary-person committed 3 years ago

36b476cc

[TensorExpr] Add support for constant tensors in tensorexpr kernel. (#56319)

Mikhail Zolotukhin committed 3 years ago

85126629

[TensorExpr] TensorExprKernel: switch type of tensors_ from Tensor to Buf. (#56318)

Mikhail Zolotukhin committed 3 years ago

dd9ef529

Allow inlining of more Tensor methods (#53905)

peterbell10 committed 3 years ago

50d4c63f

[TensorPipe] Update tensorpipe subodule + remove TP_NEW_API switch. (#56260)

Lucas Hosseini committed 3 years ago

be2a0805

[nnc] Only lower float conv2d's (#56289)

bertmaher committed 3 years ago

928a4733

Add adaptive_avgpool2d to the set of fusible ops (#56180)

Krovatkin committed 3 years ago

04e7891a

clean up unused reduction functions in THC (#56293)

ngimel committed 3 years ago

7636cb6b

A heuristic to avoid perf incompatible MKLDNN formats for binary ops (#56089)

Krovatkin committed 3 years ago

a4348358

[FX] Make shape_prop handle targets with aggregate outputs (#56221)

James Reed committed 3 years ago

d02919dd

Fix warnings in ivalue test (#56303)

davidriazati@fb.com committed 3 years ago

72a93a63

fx quant: fix subtle bug in BinaryOpQuantizeHanlder logic in matching (#56294)

vkuzo committed 3 years ago

48e675ac

Preserve node meta info in split_module (#56212)

jfix71 committed 3 years ago

5eadc243

[quant][graphmode][fx] Optimize cat (#54813)

jerryzh168 committed 3 years ago

98933866

Move graph iterator to seperate utility file (#56211)

tugsbayasgalan committed 3 years ago

cd780e1c

[JIT] Put explicit error message on class attribute accesses. (#55723)

zhxchen17 committed 3 years ago

8176ab6c

Implement faster gradcheck but not enabled for most things (#54480)

soulitzer committed 3 years ago

d312aeb6

[kineto] deprecate pthreadid (#56209)

chaekit committed 3 years ago

83cfaf1a

Fix formatting for the new language reference (#56042)

nikithamalgifb committed 3 years ago

643dd263

fixing Optional[Tensor] type in autodiff (#55565)

jjsjann123 committed 3 years ago

ce1380f9

[20/n][torch/elastic][upstream] Move torchelastic.distributed.tests to pytorch.distributed (#56215)

aivanou committed 3 years ago

d30e31cf

Newer Older