strisunshinewentingwang/pytorch

Pull Requests Commits

workaround for https://github.com/pytorch/torchdynamo/issues/1212

davidberard98 committed 1 year ago

8db17861

Re-enables some tests for linalg.det (#85141)

lezcano committed 2 years ago

2e1ec5d1

Fix behaviour of index_add / atomicAdd(bool,bool) (#85100)

lezcano committed 2 years ago

8b29b795

Added support for symbolic is_contiguous (#84829)

Chillee committed 2 years ago

4bdc0af5

[FSDP] Add `_set_flattened()`; `_is_flattened()` (#85038)

awgu committed 2 years ago

5652ab22

[vision hash update] update the pinned vision hash (#85130)

pytorchmergebot committed 2 years ago

0ec19db7

[torchdynamo hash update] update the pinned torchdynamo hash (#85129)

pytorchmergebot committed 2 years ago

b363e987

torch/jit/_trace.py in compare_outputs(original, reference, match_wha… (#84850)

SalahuddinX committed 2 years ago

647aeb83

[mergebot] rebase + merge (#85028)

clee2000 committed 2 years ago

54bccbb2

Add variable_list support to ExtractVariables struct (#84583)

steventk-g committed 2 years ago

89525cbd

TorchDynamo Remove context manager (#85124)

anijain2305 committed 2 years ago

50733c8b

Replace `expectedAlertNondeterministic` with simpler check function (#84808)

kurtamohler committed 2 years ago

95a2c3df

Remove getattr magic method from OpOverload (#85090)

ezyang committed 2 years ago

1275e2df

Performance optimizations to proxy tensor (#85049)

ezyang committed 2 years ago

00ce302c

Faster mul(sparse, sparse) with broadcasting in dense dims. (#83428)

nikitaved committed 2 years ago

d49943bd

Enable unary elementwise inplace ops for all sparse compressed layouts. (#85031)

pearu committed 2 years ago

abaf99d3

[functorch] fix jacfwd so all inputs get wrappers (#84915)

samdow committed 2 years ago

27ec195a

change the type of storage_offset to SymInt (#85102)

Krovatkin committed 2 years ago

64899c5d

[reland 2] Call jit decomp in VariableType to improve forward AD coverage (#84976)

soulitzer committed 2 years ago

7f88934a

[c10d] Ensure collectives are called with the same dtype for all tensor params. (#84664)

Rodrigo Kumpera committed 2 years ago

7dcc723d

[ONNX] Add `caffe2/python/onnx/**` to merge rule (#85118)

BowenBao committed 2 years ago

5558deac

[frontend] Print real type for Argument (#85103)

larryliu0820 committed 2 years ago

b1f5644f

Fix fetch function which breaks user code (#85099)

xuzhao9 committed 2 years ago

52a2b612

[reland] [numpy] add torch.concatenate, alias of torch.cat (#85073)

khushi-411 committed 2 years ago

2386cd29

[FSDP] Fix memory regression! (#85087)

awgu committed 2 years ago

25ecc488

Fix tied params with Fake Tensor (#85065)

eellison committed 2 years ago

4306a188

[ONNX] Enable test_custom_opsets_inverse (#85013)

Ubuntu committed 2 years ago

2e41fbc2

torch.zero_ on a sparse compressed tensor resets nnz to 0 (#85030)

pearu committed 2 years ago

a225f3cf

[ONNX] Add `third_party/onnx` to merge rule (#84715)

BowenBao committed 2 years ago

21e656b0

doc string fixed in torch.distributed.reduce_scatter (#84983)

ShisuiUzumaki committed 2 years ago

6bd7d0f8

Older