strisunshinewentingwang/pytorch

Pull Requests Commits

Olga Andreeva committed 2 years ago

b9ae5a81

Olga Andreeva committed 2 years ago

fb105ae3

Adding 2 lower precision communication hooks for FSDP

Olga Andreeva committed 2 years ago

b7c82951

Added low-precision hooks to fsdp_coom_hooks

Olga Andreeva committed 2 years ago

9221d3cf

Revert "[CI] Move CUDA-11.6 to Python-3.10 configuration (#81233)"

pytorchmergebot committed 2 years ago

c9648580

[Codemod][Format buck files with arc lint] caffe2/third_party (#81441)

jdonald committed 2 years ago

1233c3c2

[DataPipe] Fix fork/unzip with a single child (#81502)

ejguan committed 2 years ago

ccbf04dd

make clamp decomps use torch.* calls, move clamp_min/clamp_max to refs (#81619)

ngimel committed 2 years ago

50d205c5

Remove unused local variables from generator.py (#81505)

kit1980 committed 2 years ago

19a29648

Correct `super` class name (#81507)

kit1980 committed 2 years ago

e3a56938

Remove unnecessary assigment (#81498)

kit1980 committed 2 years ago

68848650

Enable Zero1's ddp_with_overlap for hpu backend (#80438)

jerome-habana committed 2 years ago

547e4997

Revert "[Py3.10] Allow floats to be imported as Long (#81372)"

pytorchmergebot committed 2 years ago

ec1b3a45

Revert "Recursively print graph module and its submodule (#81080)"

pytorchmergebot committed 2 years ago

4035a53c

create a concated LICENSE file for wheels (#81500)

mattip committed 2 years ago

37474a54

Excluded future functorch path from linters (#81563)

zou3519 committed 2 years ago

d82d2083

Add CUDA 11.7 workflows (#81095)

ptrblck committed 2 years ago

99c464ae

[torchdynamo hash update] update the pinned torchdynamo hash (#81621)

pytorchmergebot committed 2 years ago

35563f4f

Remove spurious assert (#81604)

ezyang committed 2 years ago

471397d0

Recursively print graph module and its submodule (#81080)

SherlockNoMad committed 2 years ago

fe726232

fix torch.config can't respect USE_MKLDNN flag issue (#75001)

XiaobingSuper committed 2 years ago

86b86202

Disable use_mkldnn when input is not contiguous for oneDNN (#80864)

yanbing-j committed 2 years ago

4655c3ba

[BE] Turn `_check_cuda_version` into a function (#81603)

malfet committed 2 years ago

95c148e5

[BE] Disamntle pyramid of doom in _check_cuda_version (#81602)

malfet committed 2 years ago

7e274964

[CI] Disable ios-12-5-1-x86-64 (#81612)

malfet committed 2 years ago

446833d1

[Profiler] Move Kineto activity generation into `collection.cpp` (#80796)

Taylor Robie committed 2 years ago

3289f720

[Profiler] Use parent time for implicitly finished Torch ops (#80810)

Taylor Robie committed 2 years ago

a5b5a001

Revert "[Profiler] Use parent time for implicitly finished Torch ops (#80810)"

pytorchmergebot committed 2 years ago

5392a7ec

[torchdynamo hash update] update the pinned torchdynamo hash (#81605)

pytorchmergebot committed 2 years ago

869b3ed9

[functorch hash update] update the pinned functorch hash (#81606)

pytorchmergebot committed 2 years ago

8feccedd

Older