strisunshinewentingwang/pytorch

Pull Requests Commits

move bazel and xla to unstable

clee2000 committed 2 years ago

e43a9a29

[jit] Support Awaitable type (#90863)

IvanKobzarev committed 2 years ago

2fc73622

Add CSC->BSC conversion (#92307)

alexsamardzic committed 2 years ago

53f7fb9a

Correctly restore pybind11 error_already_set (#93238)

ezyang committed 2 years ago

434eb16d

[Reland][FSDP] Do not clean FQNs for `use_orig_params=True` (#92662)

Andrew Gu committed 2 years ago

3e4d0e8d

Split the aot/dynamo TORCHDYNAMO_REPRO_AFTER cases (#93226)

ezyang committed 2 years ago

c7b03010

Update dynamic benchmark skips (#93228)

ezyang committed 2 years ago

9eb402d1

[inductor] enable more dynamic shapes tests (#93216)

nkaretnikov committed 2 years ago

04082fc0

Add vmap support for torch.index_fill (#91364)

qqaatw committed 2 years ago

5112f44d

inductor: support more conv+unary fusion (#92518)

blzheng committed 2 years ago

08035b1e

Some simple fixes (#93221)

cyyever committed 2 years ago

4d51c853

SymInt'ify view_as (#93242)

Skylion007 committed 2 years ago

e790281a

SymInt'ify reshape_as (#93241)

ezyang committed 2 years ago

3c570a2b

Apply Clang-Tidy readability-container-size-empty (#93236)

Skylion007 committed 2 years ago

0247ed27

Revert accidental change to libkineto version (#93237)

malfet committed 2 years ago

239afa0e

[Dynamo] Support out variants of ops mutate the tensors out of the function frame (#93177)

yanboliang committed 2 years ago

b3e42294

Move Sherlock to snooping dynamic shapes (#93239)

ezyang committed 2 years ago

129f1361

Set min supported Python version to 3.8 (#93155)

malfet committed 2 years ago

5976f0bd

Add two decomps for optimizer fusion (#93193)

mlazos committed 2 years ago

0dceaf07

Warn about deprecation of private decoder builtins (#93181)

Michael Gschwind committed 2 years ago

878f4f09

[Dynamo] Support enum.Enum type as dict key (#93026)

yanboliang committed 2 years ago

304d8dd6

inductor: fix inplace op's wrong lowering issue when preop is NopKernel (#92247)

XiaobingSuper committed 2 years ago

9a2becf6

inductor: make as_strided support non-contiguous input and always fix it's input layout using eager stride (#92063)

XiaobingSuper committed 2 years ago

900f8886

Add some more missing moves to aten functorch (#93098)

Skylion007 committed 2 years ago

cac1912b

[Export] Remove the concept of Scalar in export schema (#93211)

SherlockNoMad committed 2 years ago

61fd1188

[Export] Remove op filed from ex.Node schema (#93208)

SherlockNoMad committed 2 years ago

68a1065b

[vision hash update] update the pinned vision hash (#93189)

pytorchmergebot committed 2 years ago

7cc91f40

Fix endian handling in THPStorage_fromBuffer (#92834)

kiszk committed 2 years ago

cb817d61

More fixes found in tidy and libc++ (#93138)

cyyever committed 2 years ago

1e0c57b6

Fix positional issues in dedup guards (#93137)

voznesenskym committed 2 years ago

4ca511c6

Older