strisunshinewentingwang/pytorch

Pull Requests Commits

Update internal code for torch.linalg.solve

IvanYashchuk committed 3 years ago

79bf9c8b

Improve numerical stability of GroupNorm (#54921)

xiaomengy committed 3 years ago

ff15d93b

MemoryOverlap: Avoid has_storage calls (#59013)

peterbell10 committed 3 years ago

095cd6a0

[CUDA graphs] Make stream semantics of backward calls consistent with other cuda ops (ci-all edition) (#57833)

mcarilli committed 3 years ago

be038d89

Revert D28994140: [pytorch][PR] Implemented torch.cov

Mike Ruberry committed 3 years ago

92513038

Refactor SavedVariable (#59836)

Varal7 committed 3 years ago

0ceea7fa

pre compute regex and match simple signature autograd codegen 15s -> 12s (#59852)

albanD committed 3 years ago

d03ff1a1

refactor yaml loader import, no runtime change (#59850)

albanD committed 3 years ago

30a18fe3

Use C dumper if possible aten codegen 23s -> 13s (#59849)

albanD committed 3 years ago

c60d1ac9

avoid error string formatting aten codegen 28s -> 23s (#59848)

albanD committed 3 years ago

504ec301

Avoid unnecessary re-computation autograd codegen 21s -> 15s (#59847)

albanD committed 3 years ago

7143a6a1

Simplify parametrizations.SpectralNorm and improve its initialization (#59564)

lezcano committed 3 years ago

1f6e3933

Fix bad change in a CUDACachingAllocator loop (#59903)

r-barnes committed 3 years ago

10a3a3d3

Automated submodule update: FBGEMM (#59874)

facebook-github-bot committed 3 years ago

e49f0f4f

Revert D28981326: torch/lib/c10d: Use torch_check instead of throwing runtime_error

rohan-varma committed 3 years ago

3529a48e

optimize channels last for BatchNorm2d on CPU (#59286)

mingfeima committed 3 years ago

f3218568

[quant][fx] Remove extra q-dq for weight bias in normalization ops (#59882)

supriyar committed 3 years ago

864d129b

Build an -Wextra around c10 (#59853)

r-barnes committed 3 years ago

60eb22e4

make --run-specified-test-case use --include (#59704)

walterddr committed 3 years ago

e41bc31e

Fix some issues in CUDACachingAllocator (#59819)

r-barnes committed 3 years ago

cf0c4ac2

[nnc] Added a check to vectorize only those loops that are normalized. (#59423)

navahgar committed 3 years ago

b83ac0cc

[nnc] Modified vectorize API to return bool (#59422)

navahgar committed 3 years ago

30e24b2d

Remove ci/no-build (#59889)

samestep committed 3 years ago

a9e136a6

[NNC] Add python bindings for loopnest.compress_buffer (#59681)

huiguoo committed 3 years ago

f4fdc499

Give clearer lint error messages (#59876)

samestep committed 3 years ago

ee3025f7

torch/lib/c10d: Use torch_check instead of throwing runtime_error (#59684)

rohan-varma committed 3 years ago

6ea60750

Replace throw std::runtime_error with torch_check in torch/csrc/distributed (#59683)

rohan-varma committed 3 years ago

d433a55c

Fix `Vectorize<float>::trunc` on ARM platform (#59858)

malfet committed 3 years ago

9cdbddb3

[Pytorch backend delegation] Preprocess to accept (#58873)

kimishpatel committed 3 years ago

2ce21b2e

Implemented torch.cov (#58311)

heitorschueroff committed 3 years ago

23c23255

Older