Age | Commit message (Collapse) | Author | Files | Lines |
|
Summary:
This should include fix to the issue: https://github.com/NVIDIA/nccl/issues/153
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13921
Differential Revision: D13048999
Pulled By: teng-li
fbshipit-source-id: a83f3bbb004f4a4137d187a010c7ec6b48f27eeb
|
|
Summary:
Including some hang fixes. Tested locally and distributed works fine
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13353
Differential Revision: D12853714
Pulled By: teng-li
fbshipit-source-id: be72b9ffb48cffdb590e5452b0a4ec597f052685
|
|
Summary:
always build nccl from within the main cmake build, rather than via a separate invocation in build_pytorch_libs.sh. Use the existing caffe2 codepaths
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13150
Differential Revision: D12815674
Pulled By: anderspapitto
fbshipit-source-id: a710b6f242d159b9816911a25ee2c4b8c3f855aa
|
|
Summary:
- Removed the old nccl file
- Make open-source NCCL a submodule
- CMake to make NCCL itself
NCCL2 now is in the default build.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12359
Reviewed By: orionr, yns88
Differential Revision: D10219665
Pulled By: teng-li
fbshipit-source-id: 134ff47057512ba617b48bf390c1c816fff3f881
|
|
Reviewed By: The controller you requested could not be found.
fbshipit-source-id: a13075339d3a7b970e81be0b1a32a7c4c3a6c68d
|
|
Summary:
- Removed the old nccl file
- Make open-source NCCL a submodule
- CMake to make NCCL itself
NCCL2 now is in the default build.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12312
Differential Revision: D10190845
Pulled By: teng-li
fbshipit-source-id: 08d42253b774149a66919d194f88b34628c39bae
|
|
Summary:
Changes (were merged) in #8834 and #8829 (cc yf225 ) were lost in https://github.com/pytorch/pytorch/commit/9ec0a2aef4fe67e40e4f6d487c01ec614b4af11a#diff-6997846ce6daf0c271e2db9ef0508551. This PR resubmits them.
Closes https://github.com/pytorch/pytorch/pull/8948
Differential Revision: D8665760
Pulled By: SsnL
fbshipit-source-id: 15514021fa79e6b908ea665dd6cb464b3ea00ab0
|
|
|
|
|
|
* Better fix for NCCL + sccache
* Try to set NUM_JOBS to 1
* Try to fix third_party/nccl/CMakeLists.txt as well
* Pass NUM_JOBS to nccl/CMakeLists.txt
|
|
|
|
* Disable sccache when building NCCL
* Fix nccl CMakeLists.txt
|
|
* Make NCCL build select NVCC_GENCODE smarter
* add info print
* replace ; with \s
* gencode\s -> gencode=
* Don't let nccl use sccache
|
|
- gloo, pybind11, nanopb and nccl now live in third_party.
- ATen builds in aten/build rather than torch/lib/build/aten
- A bit of faffing about in the scripts was necessary, because they used to assume that everything lived in the same directory. Now you are expected to cd into the correct directory before calling one of the build functions. The actual builder script lives in tools
- Lint now just unconditionally ignores third_party, rather than enumerating folders explicitly
|