strisunshinewentingwang/pytorch

Pull Requests Commits

flash_attention_helper mitigation: pass contiguous inputs (#85135)

cpuhrsch committed 2 years ago

38778add

Increase docker build timeout (#85156)

ZainRizvi committed 2 years ago

7b3e177b

Use alias for nop decomp (#84727)

SherlockNoMad committed 2 years ago

29eba319

Rename 'torch/ao/nn/quantized._reference' to 'torch/ao/nn/quantized/reference'. (#84974)

fufeisi committed 2 years ago

d8eae628

Implement forward AD for scatter_reduce (#85000)

lezcano committed 2 years ago

d710c95c

fix half_to_float arg in *softmax decomp (#85120)

ngimel committed 2 years ago

6162a043

Re-enable fixed dynamo tests (#84969)

eellison committed 2 years ago

f37069aa

Flip fake tensors on in aot autograd (#84968)

eellison committed 2 years ago

54c9c4e7

Add warning about installing functorch via setup.py (#85095)

zou3519 committed 2 years ago

61ba1250

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

Newer Older