facebookresearch/torch_ucc issues and pull requests

#100 - Add the world size info in NCCL metadata

Pull Request - State: open - Opened by yoyoyocmu over 1 year ago - 5 comments
Labels: CLA Signed, fb-exported

#99 - use absolute path for c10d headers

Pull Request - State: closed - Opened by minsii over 2 years ago - 4 comments
Labels: CLA Signed, fb-exported

#98 - Enable capturing of comm collective parameters

Pull Request - State: closed - Opened by louisfeng over 2 years ago - 6 comments
Labels: CLA Signed, fb-exported

#97 - Separate each test set to different github action

Pull Request - State: closed - Opened by minsii over 2 years ago - 1 comment
Labels: CLA Signed

#96 - completely disable ucx when ACTIVE_SET is on

Pull Request - State: closed - Opened by minsii over 2 years ago - 1 comment
Labels: CLA Signed, fb-exported

#95 - New TORCH_UCC_BLOCKING_WAIT env variable

Pull Request - State: closed - Opened by zasdfgbnm over 2 years ago - 2 comments
Labels: CLA Signed

#94 - Fix nits from Pytorch native UCC PG comments

Pull Request - State: closed - Opened by vtlam over 2 years ago - 3 comments
Labels: CLA Signed, fb-exported

#93 - Fix compile issue with send/recv functions

Pull Request - State: closed - Opened by wesbland over 2 years ago - 1 comment
Labels: CLA Signed, fb-exported

#92 - build fail with --active-sets

Issue - State: closed - Opened by bureddy over 2 years ago - 2 comments
Labels: bug

#91 - Add clarification and fix nits in torch-ucc.

Pull Request - State: closed - Opened by vtlam over 2 years ago - 2 comments
Labels: CLA Signed, fb-exported

#90 - P2P test fails

Issue - State: open - Opened by Aidyn-A over 2 years ago

#89 - fix CI pytorch tests

Pull Request - State: closed - Opened by kingchc over 2 years ago - 1 comment
Labels: CLA Signed, fb-exported

#88 - fix comms tracing for wait

Pull Request - State: closed - Opened by kingchc over 2 years ago - 1 comment
Labels: CLA Signed, fb-exported

#87 - Use UCC Active Sets interface for pt2pt

Pull Request - State: closed - Opened by wesbland over 2 years ago - 2 comments
Labels: CLA Signed, fb-exported

#86 - remove team size 1 hack to support future object

Pull Request - State: closed - Opened by kingchc over 2 years ago - 4 comments
Labels: CLA Signed, fb-exported

#85 - fix _allgather_base to avoid garbage values

Pull Request - State: closed - Opened by kingchc over 2 years ago - 1 comment
Labels: CLA Signed, fb-exported

#84 - fix health check for CUDA with a valid device

Pull Request - State: closed - Opened by kingchc over 2 years ago - 4 comments
Labels: CLA Signed, fb-exported

#83 - generate comms trace for post-analysis and replay

Pull Request - State: closed - Opened by kingchc over 2 years ago - 2 comments
Labels: CLA Signed, fb-exported

#82 - fix health check error when multiple PGs are used

Pull Request - State: closed - Opened by kingchc over 2 years ago - 3 comments
Labels: CLA Signed, fb-exported

#81 - add allgather_base blocking wait option

Pull Request - State: closed - Opened by huiyujie over 2 years ago - 2 comments
Labels: CLA Signed, fb-exported

#80 - adding all_gather_base primitive

Pull Request - State: closed - Opened by huiyujie over 2 years ago - 2 comments
Labels: CLA Signed, fb-exported

#79 - add scatter and gather blocking wait options to Torch-UCC

Pull Request - State: closed - Opened by kingchc over 2 years ago - 1 comment
Labels: CLA Signed, fb-exported

#78 - WIP

Pull Request - State: closed - Opened by zasdfgbnm over 2 years ago
Labels: CLA Signed

#77 - Add health check support

Pull Request - State: closed - Opened by zasdfgbnm over 2 years ago - 5 comments
Labels: CLA Signed

#76 - WIP

Pull Request - State: closed - Opened by zasdfgbnm almost 3 years ago
Labels: CLA Signed

#75 - Refactor hipify setup logic for reuse

Pull Request - State: closed - Opened by minsii almost 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#74 - Hipify fbcode torch_ucc

Pull Request - State: closed - Opened by minsii almost 3 years ago - 2 comments
Labels: CLA Signed, fb-exported

#73 - enable scatter primitive

Pull Request - State: closed - Opened by kingchc almost 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#72 - Use new UCC pt2pt calls

Pull Request - State: closed - Opened by wesbland almost 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#71 - enable gather primitive

Pull Request - State: closed - Opened by kingchc almost 3 years ago - 3 comments
Labels: CLA Signed, fb-exported

#70 - Support all non-overlapping and dense tensors

Pull Request - State: closed - Opened by zasdfgbnm almost 3 years ago
Labels: CLA Signed

#69 - Run PyTorch unit tests in OSS CI

Pull Request - State: closed - Opened by zasdfgbnm almost 3 years ago - 3 comments
Labels: CLA Signed

#68 - test: always use CPU tensor with pg for result validation

Pull Request - State: closed - Opened by minsii almost 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#67 - test: refactor check funcs to support different mem type in ucc and pg

Pull Request - State: closed - Opened by minsii almost 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#66 - test: fix torch_alltoall_bench.py for ucc PG w/ cuda

Pull Request - State: closed - Opened by minsii almost 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#65 - test: lint fix and set flake8 linter for python files

Pull Request - State: closed - Opened by minsii almost 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#64 - support torch.bool datatype conversion to UCC UINT8

Pull Request - State: closed - Opened by kingchc almost 3 years ago - 4 comments
Labels: CLA Signed, fb-exported

#63 - enable alltoall primitive

Pull Request - State: closed - Opened by kingchc almost 3 years ago - 7 comments
Labels: CLA Signed, fb-exported

#62 - support reduce primitive

Pull Request - State: closed - Opened by kingchc almost 3 years ago - 5 comments
Labels: CLA Signed, fb-exported

#61 - use clang in CI tests

Pull Request - State: closed - Opened by kingchc almost 3 years ago - 6 comments
Labels: CLA Signed, fb-exported

#60 - Fail in param benchmark

Issue - State: open - Opened by avildema about 3 years ago

#59 - check for ucx mt support

Pull Request - State: closed - Opened by Sergei-Lebedev about 3 years ago - 1 comment
Labels: CLA Signed

#58 - [WIP] Add PyTorch unit tests.

Pull Request - State: closed - Opened by zasdfgbnm about 3 years ago
Labels: CLA Signed

#57 - Add a function to create ProcessGroupUCC in PyTorch

Pull Request - State: closed - Opened by zasdfgbnm about 3 years ago - 4 comments
Labels: CLA Signed

#56 - Split torch_ucc to torch_ucc and torch_ucc_oss

Pull Request - State: closed - Opened by zasdfgbnm about 3 years ago - 9 comments
Labels: CLA Signed

#55 - update ucc reduction op

Pull Request - State: closed - Opened by Sergei-Lebedev about 3 years ago - 1 comment
Labels: CLA Signed

#54 - bcopy allgatherv

Pull Request - State: closed - Opened by Sergei-Lebedev about 3 years ago - 4 comments
Labels: CLA Signed

#53 - improve error message in consensus protocol

Pull Request - State: closed - Opened by kingchc about 3 years ago - 5 comments
Labels: CLA Signed, fb-exported

#52 - fix collective_post default arg

Pull Request - State: closed - Opened by pallab-zz about 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#51 - Lower case UCC

Pull Request - State: closed - Opened by bryanmr about 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#50 - fix logger to show real line number

Pull Request - State: closed - Opened by kingchc about 3 years ago - 3 comments
Labels: CLA Signed, Merged, fb-exported

#49 - add reduce_scatter to torch_ucc

Pull Request - State: closed - Opened by pallab-zz about 3 years ago - 8 comments
Labels: CLA Signed, Merged, fb-exported

#48 - implement result interface in WorkUCC

Pull Request - State: closed - Opened by kingchc about 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#47 - Add datatype BFloat16

Pull Request - State: closed - Opened by lappazos about 3 years ago - 5 comments
Labels: CLA Signed

#46 - enable posting barrier on GPU device

Pull Request - State: closed - Opened by kingchc over 3 years ago - 4 comments
Labels: CLA Signed, Merged, fb-exported

#45 - initialize/check lib/context in CommUCC

Pull Request - State: closed - Opened by pallab-zz over 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#44 - create one commpg per pg

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 7 comments
Labels: CLA Signed

#43 - add env. variable to enable checking ucc/ucx errors

Pull Request - State: closed - Opened by kingchc over 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#42 - Add avg reduce op

Pull Request - State: closed - Opened by lappazos over 3 years ago - 7 comments
Labels: CLA Signed, Merged

#41 - support for comm size 1

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 6 comments
Labels: CLA Signed, Merged

#40 - enhance error message format and prefix

Pull Request - State: closed - Opened by kingchc over 3 years ago - 3 comments
Labels: CLA Signed, fb-exported

#39 - consensus protocol

Pull Request - State: open - Opened by Sergei-Lebedev over 3 years ago - 10 comments
Labels: CLA Signed

#38 - exchange comm id during comm create

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 8 comments
Labels: CLA Signed, Merged

#37 - override getBackendName from ProcessGroup interface

Pull Request - State: closed - Opened by kingchc over 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#36 - [POC] enhance logging and debuggability

Pull Request - State: closed - Opened by kingchc over 3 years ago - 5 comments
Labels: CLA Signed, fb-exported

#35 - Compile error

Issue - State: open - Opened by kerwenwwer over 3 years ago

#34 - fix pt2pt enqueue and hang issue

Pull Request - State: closed - Opened by kingchc over 3 years ago - 11 comments
Labels: CLA Signed, fb-exported

#33 - update oob parameters

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 2 comments
Labels: CLA Signed, Merged

#32 - add reduce_scatter CI test

Pull Request - State: closed - Opened by kingchc over 3 years ago - 5 comments
Labels: CLA Signed, fb-exported

#31 - add general memory type conversion (c10 -> UCC/UCS memtpye)

Pull Request - State: closed - Opened by kingchc over 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#30 - fix compilation error of mtype mapping

Pull Request - State: closed - Opened by kingchc over 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#29 - fix pt2pt runtime issue and add several param bench to CI test

Pull Request - State: closed - Opened by kingchc over 3 years ago - 7 comments
Labels: CLA Signed, Merged, fb-exported

#28 - retain ownership of WorkUCC in main thread to avoid deadlock

Pull Request - State: closed - Opened by kingchc over 3 years ago - 4 comments
Labels: CLA Signed, Merged, fb-exported

#27 - add param bench to CI test

Pull Request - State: closed - Opened by kingchc over 3 years ago - 15 comments
Labels: CLA Signed, fb-exported

#26 - Back out "use ucx for oob instead of store"

Pull Request - State: closed - Opened by kingchc over 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#25 - store weak pointer to ucc work

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 5 comments
Labels: CLA Signed

#24 - <DO NOT MERGE> markCompleted immediately after future_ creation for CUDA tensor

Pull Request - State: closed - Opened by kingchc over 3 years ago - 1 comment
Labels: CLA Signed, fb-exported

#23 - enable collective overlap

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 3 comments
Labels: CLA Signed

#22 - use ucx for oob instead of store

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 3 comments
Labels: CLA Signed, Merged

#21 - Adjust alltoall count

Pull Request - State: closed - Opened by lappazos over 3 years ago - 14 comments
Labels: CLA Signed, Merged

#20 - add profiling titles

Pull Request - State: closed - Opened by kingchc over 3 years ago - 2 comments
Labels: CLA Signed, Merged, fb-exported

#19 - progress context while waiting for team create

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 2 comments
Labels: CLA Signed, Merged

#18 - Post collectives from progress thread

Pull Request - State: open - Opened by Sergei-Lebedev over 3 years ago - 4 comments
Labels: CLA Signed

#17 - [DO NOT MERGE] Discussion only

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago
Labels: CLA Signed

#16 - Some tests fails

Issue - State: open - Opened by zasdfgbnm over 3 years ago

#15 - Segfault at exit

Issue - State: open - Opened by zasdfgbnm over 3 years ago

#14 - [DO NOT MERGE] Discussion only

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago
Labels: CLA Signed

#13 - Fixes segfault at exit

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago - 4 comments
Labels: CLA Signed

#12 - Use shared_ptr to store oob

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago - 2 comments
Labels: CLA Signed

#11 - Avoid using deleteKey when creating team

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago - 4 comments
Labels: CLA Signed

#10 - std::thread is used without #include <thread>

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago - 2 comments
Labels: CLA Signed, Merged

#9 - support future object in ProcessGroupUCC

Pull Request - State: closed - Opened by kingchc over 3 years ago - 6 comments
Labels: CLA Signed, Merged, fb-exported

#8 - Cleanup unused cuda ee

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago - 2 comments
Labels: CLA Signed

#7 - Save CUDA tensors in caching allocator

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago - 7 comments
Labels: CLA Signed, Merged

#6 - Set to the correct device in ProcessGroupUCC::initComm

Pull Request - State: closed - Opened by zasdfgbnm over 3 years ago - 3 comments
Labels: CLA Signed, Merged

#5 - Use pytorch nightly for CI testing

Pull Request - State: closed - Opened by srinivas212 over 3 years ago - 2 comments
Labels: CLA Signed, Merged

#4 - Remove space from copyright header

Pull Request - State: closed - Opened by srinivas212 over 3 years ago
Labels: CLA Signed

#3 - Fix Copyright Date.

Pull Request - State: closed - Opened by jladd-mlnx over 3 years ago
Labels: CLA Signed

#2 - Merge Torch-UCC into Torch_UCC repository.

Pull Request - State: closed - Opened by jladd-mlnx over 3 years ago - 2 comments
Labels: CLA Signed

#1 - Adding Contributing file

Pull Request - State: closed - Opened by facebook-github-bot over 3 years ago
Labels: CLA Signed

GitHub / facebookresearch/torch_ucc issues and pull requests