strisunshinewentingwang/pytorch

Pull Requests Commits

[dynamo] check buffers when checking accuracy (#91037)

williamwen42 committed 1 year ago

289f0643

Update patch release cherry pick condition (#90220)

albanD committed 1 year ago

17b80bfa

Bind autograd current Node for debugging purposes (#90867)

albanD committed 1 year ago

0eb45d54

use scatter_add for index_add when dim is the most inner dim (#88729)

CaoE committed 1 year ago

13dbad63

[CUDA12] Make PyTorch compatible with CUDA 12 (#91118)

jianyuh committed 1 year ago

63b8ecc4

Update dynamo xla test to make it part of the xla CI (#91130)

JackCaoG committed 1 year ago

7c58f1d4

[Checkpoint] Add test for fsdp model state saving and loading with/without resharding (#90950)

wz337 committed 1 year ago

29b119d0

fully_shard load state_dict (#90945)

rohan-varma committed 1 year ago

7330eabe

Revert "use libdevice for tanh (#90889)"

pytorchmergebot committed 1 year ago

95a115dd

[Dynamo] Fix builder for class with metaclass (#90807)

yanboliang committed 1 year ago

511fbad8

Revert "When nopython=True, Dynamo can't allow graph breaks. (#90970)"

pytorchmergebot committed 1 year ago

ef2bb9ca

Do not run inductor perf test with postnightly branch (#91133)

weiwangmeta committed 1 year ago

0f57e7f2

[Inductor] Replace graph.eliminate_dead_code() with graph.erase_node() in Permute Fusion (#91014)

jiawenliu64 committed 1 year ago

857ed2d7

[inductor] Rewrite Triton templates + epilogue fusion (retry) (#91105)

jansel committed 1 year ago

d6dd2e97

Removing invalid git option when updating submodules (#91132)

atalman committed 1 year ago

3bd37ff2

use libdevice for tanh (#90889)

ngimel committed 1 year ago

01488091

Fix non-existing parameters in docstrings in benchmarks (#91115)

kit1980 committed 1 year ago

30edd39b

Fix non-existing parameters in docstrings in misc places (#91121)

kit1980 committed 1 year ago

99bd8d12

Fix terminology within `linalg.slogdet` docs (#91129)

lezcano committed 1 year ago

0210d508

[Quant] lower fused LinearTanh for onednn backend (#89188)

Xia-Weiwen committed 1 year ago

a5eb564b

TorchDynamo: set output stride using eager output for cat (#89477)

XiaobingSuper committed 1 year ago

666d2180

Add catch socket.gaierror for _matches_machine_hostname (#91119)

cz-37 committed 1 year ago

b309599d

[MPS] Fix the assert in Garbage Collector (#91106)

razarmehr committed 1 year ago

ebea45fe

[pytorch profiler] Add step tracker logic to handle multiple sources of step increments (#90880)

briancoutinho committed 1 year ago

1d3e7fcc

[torch.func] Setup torch.func, populate it with all transforms (#91016)

zou3519 committed 1 year ago

41846e20

Stop using :attr: in functorch docs (#91015)

zou3519 committed 1 year ago

cad1ce61

When nopython=True, Dynamo can't allow graph breaks. (#90970)

wschin committed 1 year ago

7e9bf2ed

Autocast support for scaled_dot_product_attention (#91066)

mikekgfb committed 1 year ago

d1772aff

Propagate guard failures to userland (#91053)

ezyang committed 1 year ago

fadf2226

Delete dynamic_propagation config (#91040)

williamwen42 committed 1 year ago

7bc3467f

Newer Older