strisunshinewentingwang/pytorch

Pull Requests Commits

Refactor subgraph merging (#44238)

Elias Ellison committed 3 years ago

0137e364

Add operator to compute the equalization scale (#45096)

Haixin Liu committed 3 years ago

1539d4a6

Add architectural support for multi-GPU. (#44059)

Ashkan Aliabadi committed 3 years ago

5a593306

Minor touchups. (#44317)

Ashkan Aliabadi committed 3 years ago

6311c5a4

Grammatically updated the tech docs (#45192)

Vishal487 committed 3 years ago

b84dd771

[caffe2] adds hypothesis test for queue ops cancel (#45178)

dahsh committed 3 years ago

cd7a6822

[JIT] Specialize AutogradZero: merge AutogradAnyNonZero and Not(AutogradAnyNonZero) checks into one. (#44987)

Mikhail Zolotukhin committed 3 years ago

71e6ce66

[caffe2] adds Cancel to SafeDequeueBlobsOp and SafeEnqueueBlobsOp (#45177)

dahsh committed 3 years ago

cbe1eac1

Make ddp_comm_hook_wrapper a private method. (#44643)

Yi Wang committed 3 years ago

022ba5a7

[Caffe2] Fix LayerNormOp when batch_size == 0. (#45250)

xiaomengy committed 3 years ago

e2bcdc7b

Run pytorch_core CUDA tests on GPU using TPX

malfet committed 3 years ago

c3a5aed5

add rocm 3.8 to nightly builds (#45222)

jeffdaily committed 3 years ago

c211a910

Revert D23753711: [pytorch][PR] Add foreach APIs for binary ops with ScalarList

Xinyu Li committed 3 years ago

26001a23

added rocm 3.8 docker image (#45205)

KyleCZH committed 3 years ago

c79d4930

Adjust TF32 tests (#44240)

zasdfgbnm committed 3 years ago

3f5eee66

[hotfix] typo in NaiveConvolutionTranspose2d.cu (#45224)

Rong Rong committed 3 years ago

b8eab8cd

Add a warning log when there is high skew of uneven inputs in DDP training (#45238)

rohan-varma committed 3 years ago

e57a0811

Moves prim ops from C10 back to JIT (#45144)

raziel committed 3 years ago

2b38c09f

replace timer test with a mocked variant (#45173)

Taylor Robie committed 3 years ago

8507ea22

Bump up NCCL to 2.7.8 (#45251)

mrshenli committed 3 years ago

bfdf4323

adding a test for ddp save()/load() (#44906)

bdhirsh committed 3 years ago

5195d727

renaming TestDdpCommHook class so it doesn't get picked up as a test by pytest (#44905)

bdhirsh committed 3 years ago

f9ae296a

add skip_if_rocm to all requires_nccl tests (#45158)

Rong Rong committed 3 years ago

bc591d76

Add foreach APIs for binary ops with ScalarList (#44743)

izdeby committed 3 years ago

71d1b5b0

Enable torch.tensor typechecks (#45077)

Rong Rong committed 3 years ago

bea7901e

Deprecate old fft functions (#44876)

peterbell10 committed 3 years ago

dc67b47b

gtest-ify JIT tests, through the letter c (#45249)

suo committed 3 years ago

6d21d5f0

Sparse softmax support (CUDA) (#42307)

aocsa committed 3 years ago

29dc3c5e

[ONNX] Update ONNX docs for release (#45086)

neginraoof committed 3 years ago

b3d7c2f9

[TensorExpr] Fix min and max for integral inputs in CUDA backend (#44984)

asuhan committed 3 years ago

3dd0e362

Newer Older