strisunshinewentingwang/pytorch

Pull Requests Commits

change test behavior for mha hitting fastpath to accomodate fastpath disabled now with attn_mask

erichan1 committed 2 years ago

b8dad7d5

erichan1 committed 2 years ago

821e1fcc

disable attn_mask for mha because crash

erichan1 committed 2 years ago

09d5c181

Fix typos under torch/distributed directory (#98225)

kiszk committed 2 years ago

6514d71a

remove typed StorageImpl::unsafe_data() (#98218)

mikey dagitses committed 2 years ago

3af02283

Increment pending forwards after invocation (#98101)

eellison committed 2 years ago

a3365e1d

[SyncBatchNorm] Support running with low precision parameters (#98332)

Andrew Gu committed 2 years ago

3686416a

Extend TensorImpl with BackendMeta (#97429)

ppiskorski committed 2 years ago

2d9b2bcf

Revert "[pt2][inductor] add `triton.__verison__` as cache key, update cache layout (#98010)"

pytorchmergebot committed 2 years ago

dd503376

[CI] Mark mobilenet_v3_large as nondeterministic (#98314)

desertfire committed 2 years ago

bd6db542

[ROCm] Enable test_filtering_env_var (#84100)

jaglinux committed 2 years ago

ecf08a0f

Set number of threads to be 1 for ARM (#97482) (#98267)

jasonjk-park committed 2 years ago

51a978fe

[FSDP][Docs] Add warning about forward saving param refs (#98320)

Andrew Gu committed 2 years ago

aaae5887

[FSDP] Only move current FSDP's states to GPU during init (#98319)

Andrew Gu committed 2 years ago

66d07e3b

[FSDP] Add skip writeback check gated by env var (#98300)

Andrew Gu committed 2 years ago

d7156175

Support Modules with custom __getitem__ method through fallback (#97932)

mreso committed 2 years ago

96595617

[EASY] Teach requires_bwd_pass how to interpret int. (#98312)

ezyang committed 2 years ago

05791174

[Dynamo] graph break when calling resize_() on graph input (#98279)

yanboliang committed 2 years ago

fd0be80d

[EASY] Handle new inference csv from CI (#98294)

ezyang committed 2 years ago

3c36f82f

Propogate dynamo shape_env to make_fx (#96437)

tugsbayasgalan committed 2 years ago

75ac6fdc

[pt2][inductor] add `triton.__verison__` as cache key, update cache layout (#98010)

nmacchioni committed 2 years ago

0eab3ab5

[kineto] make input shape collection opt-in for on-demand tracing (#746) (#97917)

chaekit committed 2 years ago

9ddd97e1

Fix ideep submodule (#98305)

Rodrigo Kumpera committed 2 years ago

b04f8636

add Half support for sigmoid on CPU (#96077)

mingfeima committed 2 years ago

34c7adf1

Deduplicate pointers to manually free (#98097)

eellison committed 2 years ago

89dc87a2

Revert "Add arm tests to mps workflow (#97279)"

pytorchmergebot committed 2 years ago

a52cf339

Introduce torch.onnx.dynamo_export API (#97920)

abock committed 2 years ago

558e5a24

[FSDP][Easy] Remove unused `requires_grad_mask` (#98299)

Andrew Gu committed 2 years ago

fe9da298

Cleanup redundant CI jobs (#98044)

huydhn committed 2 years ago

4934dde3

[FSDP] Skip `_use_sharded_views()` for `SHARD_GRAD_OP` (#98250)

Andrew Gu committed 2 years ago

10271a60

Older