strisunshinewentingwang/pytorch

Pull Requests Commits

[kineto] Enabled dynamic linking with CUPTI (ci-all)

ilia-cher committed 3 years ago

4cdf9338

[profiler] Add profiler fallback (#57612)

Ilia Cherniavskii committed 3 years ago

2370d8c4

[c2] Fix flaky test_spatial_bn_multi_batch_grad

valentinandrei committed 3 years ago

da06ae73

[JIT] Lazily initialize aliasDb in DCE (#56649)

desertfire committed 3 years ago

eb6445a9

Improve BatchNorm1d performance (CUDA) (#57034)

peterbell10 committed 3 years ago

b2936ad8

[paramcomms] support for in and out split sizes (#57709)

chaekit committed 3 years ago

1101a5f6

Port ceil to structured (#57589)

Freey0 committed 3 years ago

ebd2c0a4

Port sign to structured (#57588)

Freey0 committed 3 years ago

ccbaa5fb

Port floor to structured (#57587)

Freey0 committed 3 years ago

3c0d22fa

TensorIterator: documentation on the order of creation (#57550)

shamanDevel committed 3 years ago

d83d1d37

Revert D28242069: Add cuSOLVER path for torch.linalg.lstsq

samestep committed 3 years ago

72ebdd68

Add result() to ProcessGroupGloo::AsyncWork's (#57565)

agolynski committed 3 years ago

dc06f524

Update internal code for torch.lu_solve (#56611)

IvanYashchuk committed 3 years ago

a7ba0f08

Torchelastic: populate __init__.py with failover documentation

aivanou committed 3 years ago

cb7197ce

Fixed the error in conv1d example (#57356)

harishsdev committed 3 years ago

ad31aa65

Add note about improved vmap prototype to vmap docstring (#57677)

zou3519 committed 3 years ago

52ac015d

[AutoAccept][Codemod][FBSourceClangFormatLinter] Daily `arc lint --take CLANGFORMAT`

CodemodService FBSourceClangFormatLinterBot committed 3 years ago

f1a62264

Revert D28154522: Add call_kwargs(args, kwargs) method to torch::deploy api

Mike Ruberry committed 3 years ago

40cb55f9

Add cuSOLVER path for torch.linalg.lstsq (#57317)

IvanYashchuk committed 3 years ago

7b31d426

Add CUDA support for torch.ormqr (#57316)

IvanYashchuk committed 3 years ago

35fab44e

Port CPU torch.ormqr to ATen (#57315)

IvanYashchuk committed 3 years ago

59d794b2

[pytorch][nnc] mobile nnc backend skeleton (#56852)

ljk53 committed 3 years ago

b4a098f1

[pytorch][nnc] protocol classes to persist the context for compiled functions (#56851)

ljk53 committed 3 years ago

d82333e9

Fix internal assert in CUDA caching allocator when trying to allocate ~2^64 memory (#57571)

zasdfgbnm committed 3 years ago

db7b3135

Make RRefContext get devices from RPC agent when creating OwnerRRef (#57443)

lw committed 3 years ago

7d4121d1

Replace DeviceIndexes with Devices in RRefs (#57442)

lw committed 3 years ago

7ffadf6e

Make DataPtr extraction in CUDAFuture faster for Python values (#56918)

lw committed 3 years ago

8e9bbd31

Ensure devices are preserved when forwarding between futures (#57432)

lw committed 3 years ago

69de4940

Avoid re-extracting DataPtrs when forwarding values between Futures (#57433)

lw committed 3 years ago

12926023

[NNC] Add support for computing conv with dynamic shapes (#57514)

navahgar committed 3 years ago

1f178de8

Older