strisunshinewentingwang/pytorch

Pull Requests Commits

[typing] suppress errors in `fbcode/caffe2` - batch 2

generatedunixname89002005307016 committed 4 years ago

c4f50162

Fix SIGSEGV in CudaIPCTypes.cpp. (#53080)

pritamdamania committed 4 years ago

8533a485

TST Replaces assertEqualIgnoreTypes with assertEqual in test_indexing (#53115)

thomasjpfan committed 4 years ago

dc070605

[Gradient Compression] Introduce fp16_compress_wrapper in ddp_comm_hooks.rst (#54052)

Yi Wang committed 4 years ago

4b00bce1

[PyTorch Mobile] Dedup method names in bytecode serialization (#53677)

iseeyuan committed 4 years ago

524cb0a5

Delete defunct ComplexCPU/ComplexCUDA dispatch keys (#54013)

ezyang committed 4 years ago

282eefeb

Make storage access error NotImplementedError (#53972)

ezyang committed 4 years ago

48784156

Compute type_equal() without reference to backend() (#53823)

ezyang committed 4 years ago

d47fd3df

Also propagate storage_access_should_throw_ when copying tensor metadata (#53816)

ezyang committed 4 years ago

3c457043

[PyTorch][JIT] Audit interpreter for extra copies (#54029)

swolchok committed 4 years ago

665d5e2a

various doc building cleanups (#53851)

mattip committed 4 years ago

ae154a8c

[complex] torch.lerp: complex autograd support (#53689)

kshitij12345 committed 4 years ago

aa8714df

Don't actually print anomalies in TTRR (#54078)

samestep committed 4 years ago

c0fafcc7

Faster backwards for cumsum and cumprod (#53711)

lezcano committed 4 years ago

1f5b9170

enable sc1090 and sc1091 (#54069)

Rong Rong (AI Infra) committed 4 years ago

6332fd62

fix mkldnn_add in-place behavior (#51687)

msy-kato committed 4 years ago

2c4a6458

[RELAND] [CUDA graphs] Private mempools for CUDA graphs (#54038)

Michael Carilli committed 4 years ago

b27e678d

remove aliasMultinomial decode from TH and THC (#52585)

XiaobingSuper committed 4 years ago

bea3cb70

[MacOS] Cross compile stub when building for M1 on x86 (#54046)

malfet committed 4 years ago

e8e570e9

Pass Scalar by reference (#53583)

wenleix committed 4 years ago

2ecb2c79

Treat Scalar parameter as if it is constant (#53582)

wenleix committed 4 years ago

4dd1c72d

OneDNN MaxPooling: reduce memory use for inference path (#52728)

XiaobingSuper committed 4 years ago

603097be

[PyTorch Mobile] Add module size to logged metadata (#53578)

dhruvbird committed 4 years ago

2c557970

Test forward reference annotations (#53713)

Ansley Ussery committed 4 years ago

08f04c0d

Disabling dispatch to OneDNN for group convolutions when groups size = 24 * n (#53991)

VitalyFedyunin committed 4 years ago

ce2f7183

[DDP] remove dedupe check in reducer (#53919)

rohan-varma committed 4 years ago

f52a3bd6

[package] autoformat (#53783)

suo committed 4 years ago

8c2c9450

Fix sharding algo + test it (#53942)

janeyx99 committed 4 years ago

ee350608

[4/n][torch/elastic][upstream] Move torchelastic/metrics to torch/distributed/elastic/metrics (#53870)

aivanou committed 4 years ago

e91aeb04

Fix doc (#53996)

nikithamalgi committed 4 years ago

b9fdf721

Older