summaryrefslogtreecommitdiff
path: root/torch/lib
AgeCommit message (Expand)AuthorFilesLines
2018-11-06Consolidate argument checkers (#13623)Pieter Noordhuis2-119/+131
2018-11-05Replaces c10d's CUDAEvent with ATen's (#13464)mruberry12-211/+64
2018-11-05Add new style broadcast support in c10d/gloo (#13497)Pieter Noordhuis1-36/+149
2018-11-05Add new style allreduce support in c10d/gloo (#13426)Pieter Noordhuis1-35/+184
2018-11-05Add reduce support in c10d/gloo (#13425)Pieter Noordhuis1-3/+105
2018-11-05Add allgather support in c10d/gloo (#13424)Pieter Noordhuis2-3/+142
2018-11-05Add gather support in c10d/gloo (#13423)Pieter Noordhuis1-4/+95
2018-11-05Add scatter support in c10d/gloo (#13422)Pieter Noordhuis3-8/+255
2018-11-03minor build fixes for incremental builds (#13293)Soumith Chintala1-1/+1
2018-11-01Replace tmpnam usage (#13289)Peter Goldsborough2-8/+16
2018-10-31Revert "Revert D12858091: [pytorch][PR] restore USE_C10D_NCCL" (#13407)Anders Papitto1-0/+4
2018-10-31Revert D12858091: [pytorch][PR] restore USE_C10D_NCCLWill Feng1-4/+0
2018-10-31restore USE_C10D_NCCLAnders Papitto1-0/+4
2018-10-31Replace CUDA-specific set_index(_from) method from DeviceGuard with set_devic...Edward Yang6-29/+31
2018-10-31Delete DeviceGuard(int64_t) constructor. (#13232)Edward Yang3-5/+7
2018-10-29absorb nccl (#13150)Anders Papitto3-53/+9
2018-10-29Delete default constructor from CUDAStream. (#13021)Edward Yang4-15/+16
2018-10-28Fix move constructor on c10d::CUDAEvent (#13183)Edward Yang1-3/+3
2018-10-26reduce Device to 32bits (#12767)Roy Li1-2/+4
2018-10-25clang-format on c10d and THD (#13138)Teng Li48-1806/+2280
2018-10-25Absorb shmAnders Papitto2-5/+18
2018-10-24Absorb c10d into the main cmake buildAnders Papitto2-29/+23
2018-10-24absorb THD into main cmake build (#12775)Anders Papitto1-13/+17
2018-10-24Make c10d support MPICH and further (#13083)Teng Li1-1/+4
2018-10-24link libgloo_cuda.a explictly from setup.py (#12951)Anders Papitto1-4/+0
2018-10-24Rename createCUDAStream() to getStreamFromPool() (#12940)Edward Yang4-4/+4
2018-10-19Delete THCStreamGuard in favor of CUDAGuard, also c10d code cleanup (#12849)Edward Yang14-180/+80
2018-10-18Link libshm against c10 (#12802)James Reed1-1/+1
2018-10-15Move exception to C10 (#12354)Yangqing Jia1-1/+1
2018-09-25Move Factory functions from Type to TypeExtendedInterface. (#12025)Gregory Chanan1-1/+1
2018-09-24Fix return temporary as reference in MPI backend (#11947)Pieter Noordhuis2-10/+28
2018-09-24Get rid of most usages of Type.tensor. (#12002)Gregory Chanan2-3/+3
2018-09-21Revert D9985212: [pytorch][PR] [minor] remove a remaining todo line deletion ...Yangqing Jia1-0/+2
2018-09-21remove a remaining todo line deletion in THD cmake (#11920)Yangqing Jia1-2/+0
2018-09-20Defer lazyInitCUDA() until needed (#11893)Pieter Noordhuis2-18/+6
2018-09-20Support torch.distributed.barrier in gloo backendPieter Noordhuis3-15/+59
2018-09-17Reorder statements for readability (#11764)Pieter Noordhuis1-16/+13
2018-09-14Add message tag parameter to send/recvPieter Noordhuis8-28/+59
2018-09-13Use MPI_Isend/MPI_Irecv to back send/recv (#11630)Pieter Noordhuis7-61/+156
2018-09-12Make c10d test work on CPU only build (#11567)Teng Li3-0/+18
2018-09-11Make C10d support CPU only build (#11513)Teng Li4-45/+124
2018-09-07bumping up the default store timeout (#11409)Teng Li1-1/+1
2018-09-07Support send/recv for the gloo process group (#11387)Pieter Noordhuis2-9/+184
2018-09-07Remove hardcoded system path from CMAKE_MODULE_PATH (#11386)Pieter Noordhuis1-1/+0
2018-09-06Adding setTimeout option in Store (#11265)Teng Li8-4/+46
2018-09-05Fix some more warnings (#11257)Tongzhou Wang2-2/+2
2018-08-31PT1 Release Milestone No.2 MPI Group Support with all tests passed (#11128)Teng Li9-51/+171
2018-08-30Delete a bunch of uses of getType in favor of TensorOptions.Edward Yang3-14/+8
2018-08-30Creates stream pool (#9938)mruberry1-1/+1
2018-08-29PT1 Distributed Release MileStone No.1 - Completed Distributed Package and CI...Teng Li1-8/+9
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-29 14:09:21 +0000