strisunshinewentingwang/pytorch

Pull Requests Commits

[dynamo][guards] Use dict for storing weakrefs (#107645)

anijain2305 committed 1 year ago

8c62f01c

Fix deepcopy for tensor with MTIA device key. (#107427)

egienvalue committed 1 year ago

221daeb1

Use TORCH_SYM_CHECK for check_size_nonnegative on SymIntArrayRef (#107785)

ezyang committed 1 year ago

42b6ba34

[2/N][DeviceMesh] Overriding __getitem__ for DeviceMesh to support Mesh Slicing (#107730)

wz337 committed 1 year ago

cdd0821f

Expose torch.export.constrain_as_{size,value} APIs (#107735)

gmagogsfm committed 1 year ago

652ccfad

Remove unnecessary import in python_variable.cpp (#107794)

janeyx99 committed 1 year ago

9d23b8b3

[dynamo] Treat monkey patched .forward as dynamic (#107104)

jansel committed 1 year ago

79b3a9f9

Revert the removal of a SampleInput for gather (#107776)

lezcano committed 1 year ago

977aba7c

[allocator] register oom observers on every device (#107399)

zdevito committed 1 year ago

c9b5e9d7

[memory snapshot] add 'address' key to block (#107171)

zdevito committed 1 year ago

cc54448a

[FSDP] Enable async all-reduce for HSDP (#106080)

awgu committed 1 year ago

2b964d6e

[FSDP] Break up `_post_backward_hook` into smaller funcs (#106068)

awgu committed 1 year ago

50e13786

torch._numpy: keep f16 CUDA tensors in f16 where possible (#107768)

ev-br committed 1 year ago

55d6b801

pt2: make aot_eager backend handle basic float8 operations (#107783)

vkuzo committed 1 year ago

61fe49b8

[ONNX] More debug logging from fx to onnx (#107654)

BowenBao committed 1 year ago

5b632bf7

[ONNX] Clean up diagnostic rules (#107653)

BowenBao committed 1 year ago

bb1852fb

[ONNX] Enclose package info for modules exported as local functions (#107409)

BowenBao committed 1 year ago

c3c1b68a

[ONNX] Re-purpose 'name' field of GraphProto (#107408)

BowenBao committed 1 year ago

7a8db57e

Back out "[inductor] make thread order consistent with loop order (#106827)" (#107796)

jackiexu1992 committed 1 year ago

398f4ae4

fix pad_sequence docstring (#107669)

kato8966 committed 1 year ago

f7a51c42

Skip the extra copy operation in broadcast_object_list if tensor_list has only one element (#107509)

Codle committed 1 year ago

42738c56

Revert "reseed all Generators in Dataloader's _worker_loop() -- via GC (#107131)"

pytorchmergebot committed 1 year ago

ecde6226

[inductor] Separate to_{dtype,device} from lowering to avoid copying (#107640)

peterbell10 committed 1 year ago

3f2ecf77

test_memory_format test now passes on rocm (#107696)

pragupta committed 1 year ago

3022a395

[CI] Delete .github/ci_commit_pins/huggingface.txt (#107729)

desertfire committed 1 year ago

469e7479

[CODEOWNERS] Add wz337 as a reviewer for Distributed Package and Distributed Tests. (#107747)

wz337 committed 1 year ago

9f5c7058

fix type check of overflow (#107579)

FFFrog committed 1 year ago

6f0d0b38

Disable nn.MHA fastpath for floating point masks (#107641)

mikaylagawarecki committed 1 year ago

48b1208e

[dynamo] Wrap ndarray dunder methods (#107689)

lezcano committed 1 year ago

207b06d0

[dynamo] Fix ndarray.__pow__ (#107746)

lezcano committed 1 year ago

b5c90ba7

Newer Older