- 02 Apr, 2021 1 commit
-
-
msbaines authored
NCCL all_to_all is now supported in PyTorch (since v1.8.0) Fixes: #548
-
- 11 Jan, 2021 1 commit
-
-
Benjamin Lefaudeux authored
* tentatively fixing the cpu version of circleci jobs, now pipe tests are the last ones standing * fixing oss backcompat, trying to fix rpc in old pytorch also * fixing the file based init in torch 1.5
-
- 28 Dec, 2020 1 commit
-
-
Benjamin Lefaudeux authored
* file based dist init * nicer handling of broken world sizes vs. number of available GPUs, do not break but warn out
-
- 11 Nov, 2020 2 commits
- 10 Nov, 2020 1 commit
-
-
Tom Birch authored
Adds support for: * Reused layers (e.g. for weight sharing) * Lazily-constructed layers * Single-process control via PipeRPCWrapper * PipelineStyle.AsyncScheudle, which lays the foundation for asynchronous pipeline work by introducing an event loop for each rank/worker to process either activations or gradients as they arrive Also added examples for multi-process and PipeRPCWrapper
-
- 23 Oct, 2020 1 commit
-
-
msbaines authored
-
- 21 Oct, 2020 1 commit
-
-
msbaines authored
-
- 17 Oct, 2020 1 commit
-
-
msbaines authored
-
- 16 Oct, 2020 2 commits
- 14 Oct, 2020 1 commit
-
-
msbaines authored
-
- 08 Oct, 2020 1 commit
-
-
msbaines authored
Currently only implemented for a single process and expert.
-
