strisunshinewentingwang/pytorch

Pull Requests Commits

[FSDP] Include module classes in `ModuleWrapPolicy.__repr__` (#89058)

Andrew Gu committed 3 years ago

ce17bb95

[FSDP] Limit all gather after pre-unshard (#89057)

Andrew Gu committed 3 years ago

c8aaad04

[Checkpoint][2D][5/N] Add checkpoint_utils for distributed checkpoint to testing/_internal/distributed/ (#89873)

wz337 committed 3 years ago

56b3ad78

[FSDP] Remove unneeded stream sync from `clip_grad_norm_()` (#89308)

Andrew Gu committed 3 years ago

be80b72a

Generator of tensor inputs with variable layout and structure (batch/non-batch, hybrid/non-hybrid, block/non-block) (#88914)

pearu committed 3 years ago

90bed887

Enable rsqrt (#89771)

eellison committed 3 years ago

275ade63

add env/config flag to disable dynamo (#89828)

mlazos committed 3 years ago

2d32e5dd

[functorch] Move `cond.py` to `_cond.py` and expose `cond()` under functorch.experimental.control_flow. (#89819)

zhxchen17 committed 3 years ago

a70082a8

[FSDP][Easy] Remove outdated TODO (#89217)

Andrew Gu committed 3 years ago

d1760d7a

Make fake tensors preserve dense strides in type conversion (#89803)

eellison committed 3 years ago

1a33b7cb

[checkpoint] Improve test (test_nested_dict.py) (#89854)

wz337 committed 3 years ago

9c8a94bf

Fix typo in filesystem.py (#89849)

wz337 committed 3 years ago

cefece37

[dashboaard] Fix flag compilers (#89853)

anijain2305 committed 3 years ago

5a79144a

[CI] Add TorchTrition conda packages (#89841)

malfet committed 3 years ago

59a2fe74

[Caffe2] Fix merge logic bug (#89551)

scramsby committed 3 years ago

24b3b73c

Remove beauby and dzdang from CODEOWNERS (#89811)

kit1980 committed 3 years ago

55789b40

[inductor] Add aten._native_batch_norm_legit to decomposition (#89843)

desertfire committed 3 years ago

693135a9

Update code style for optimizer code (#89862)

mlazos committed 3 years ago

3d47c74c

[quant][docs] Move some of the descriptions out of codeblock (#89795)

jerryzh168 committed 3 years ago

8ca09dda

Enable instance norm running mean test (#89793)

janeyx99 committed 3 years ago

fcb5d6e7

[FSDP] Another fix for `DTensor`, `use_orig_params=True` (#89845)

Andrew Gu committed 3 years ago

c599cf24

replace double transpose with single permute in nn.f.mha (#89847)

drisspg committed 3 years ago

b9afa928

Stream actually overrides __new__ so we need to patch it as well (#89592)

albanD committed 3 years ago

8713119c

Move gpu slow tests to sm86 (#87880)

davidberard98 committed 3 years ago

a029ec2c

Deprecating DataPipes (#89794)

ejguan committed 3 years ago

991028cd

Don't unsafely clone autograd meta (#89720)

ezyang committed 3 years ago

6c1fb3f2

Fix CopySlices logic to ensure wrapped node runs properly. (#89812)

albanD committed 3 years ago

02e2eaa9

[test_nn] split multihead_attention from test_nn (#89748)

kshitij12345 committed 3 years ago

8314d403

[Quant][docs] Use get_default_qconfig_mapping (#87299)

andrewor14 committed 3 years ago

fb47a669

[Quant][fx][bc-breaking] Remove backend_config_utils.py (#89810)

andrewor14 committed 3 years ago

2bce6d09

Newer Older