strisunshinewentingwang/pytorch

Pull Requests Commits

Fix multi gpu test break on Windows

skyline75489 committed 3 years ago

Verified 0c4b148b

Revert D28387765: Add forward AD gradcheck

Mike Ruberry committed 3 years ago

2d7d6922

Revert D28387767: Add forward AD test for op info

Mike Ruberry committed 3 years ago

f88297c6

Allow instruction counting to use shared memory as a staging ground. (And a couple other tweaks.) (#56711)

Taylor Robie committed 3 years ago

87f7fdfd

Revert D28387764: Codegen inplace forward AD formula from out of place one if needed

ezyang committed 3 years ago

066e7699

Enabled `roll` & `diag` for BFloat16 dtype on CUDA (#57916)

imaginary-person committed 3 years ago

ce1a8620

Enable lerp for BFloat16 on CUDA (#57907)

imaginary-person committed 3 years ago

f9aa6b24

Enable `ceil`, `floor`, `frac`, `round` & `trunc` for BFloat16 on CUDA (#57910)

imaginary-person committed 3 years ago

e6d8f455

Enable atan2 & hypot for BFloat16 on CUDA (#57905)

imaginary-person committed 3 years ago

c4a486f4

Add LowerSimpleTuples for freeze tuples (#57915)

FloydHsiu committed 3 years ago

f4a5730a

[torch/elastic] Add logging to the sanitize function of RendezvousStateHolder (#58169)

cbalioglu committed 3 years ago

f0a55007

Codegen inplace forward AD formula from out of place one if needed (#57767)

albanD committed 3 years ago

22799621

Add forward AD test for op info (#57701)

albanD committed 3 years ago

26b6d044

Add forward AD gradcheck (#57633)

albanD committed 3 years ago

647282cb

Update docs for get_future support (#58107)

agolynski committed 3 years ago

bc30c316

move `flatten_dense_tensors` and `unflatten_dense_tensors` to Native (#58006)

eqy committed 3 years ago

645a5f70

fix lint (#58203)

suo committed 3 years ago

f1ac9b65

[torch/elastic] Update the rendezvous docs (#58160)

cbalioglu committed 3 years ago

028f2f62

[torch/elastic] Revise distributed run script (#58159)

cbalioglu committed 3 years ago

ae63b1d1

[reland] make ddp logging api to be private (#58089)

zhaojuanmao committed 3 years ago

166a8df6

enable deterministic path for index_copy_cuda with index_put (#58144)

yuguo68 committed 3 years ago

8a450067

[package] Add an intern keyword (#57341)

suo committed 3 years ago

01d0eb9d

Combine backtrace print into one string to avoid interleaving. (#56961)

pritamdamania committed 3 years ago

d230045f

OpInfo: narrow (#58082)

kshitij12345 committed 3 years ago

d09abf00

enable support for nested containers in `torch.testing.assert(equal|close)` (#57270)

pmeier committed 3 years ago

9148f19e

Infer types for arguments of methods not invoked directly by monkeytype (#57202)

nikithamalgi committed 3 years ago

9063cb0a

[ONNX] Handle PackedParams inputs for _propagate_and_assign_input_shapes (#56449) (#57079)

neginraoof committed 3 years ago

1de3525c

Back out "[PyTorch Edge] Reuse constant table from ts in bytecode" (#58099)

cccclai committed 3 years ago

3d5bb710

Port threshold to structure (#57810)

liuyuanqiang@bytedance committed 3 years ago

85d64648

Delete move constructor on TensorImpl (#58048)

ezyang committed 3 years ago

82b2013e

Older