Only call Init Distributed if required #734
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
sdtblck
added a commit
to EleutherAI/DeeperSpeed
that referenced
this pull request
Feb 11, 2021
* Dist testing backend fixes, etc. (deepspeedai#708) * set_batch_fn and remove old sanity check (deepspeedai#712) * properly set engine.local_rank if it's set to -1 * Add executable permission to `ds_elastic` and `ds_report` in `bin`. (deepspeedai#711) * Add executable permission to `ds_elastic` and `ds_report` in `bin`. * Automatic `ds_elastic` formatting Co-authored-by: Jeff Rasley <jerasley@microsoft.com> * local rank of -1 means not set (deepspeedai#720) * bump to 0.3.11 * [launcher] look ma, no more zombies (deepspeedai#714) Co-authored-by: Jeff Rasley <jerasley@microsoft.com> * Improve starred expressions (deepspeedai#696) * Improve starred expressions `deepspeed/profiling/flops_profiler/profiler.py` uses starred expressions that are no longer valid with [PEP 617][1]. The new Python parser is in 3.9, and this change allows DeepSpeed to run with the newest Python version. I have not checked all locations that has this issue. However, this change allows me to run simple examples. [1]: https://www.python.org/dev/peps/pep-0617/ * Match style for "Improve starred expressions", although readability suffers The style guide might need to be updated for this new use case of expressions. Python [Issue 40631][1] includes more discussion on the change. [1]: https://bugs.python.org/issue40631 Co-authored-by: Cheng Li <pistasable@gmail.com> * Fixed typo in Readme. (deepspeedai#737) * 1bit_adam dependencies (deepspeedai#742) * Clickable screenshots (deepspeedai#746) * Fix docstring * Make screenshots clickable for easier viewing * Add flops profiler tutorial (deepspeedai#682) * work on flops profiler tutorial * update flops profiler tutorial * add flops profiler tutorial and fix names * work on flops profiler tutorial * update flops profiler tutorial * add flops profiler tutorial and fix names * fix tailing ws * fix names * remove multistep profiling and update docs * fix cases where functionals and submodules coexist in a parent module, update readme * fix typo * always invoke post hook function * fix module flops sum and update tests * update tutorial * Only initialize distributed if required (deepspeedai#734) Co-authored-by: Jeff Rasley <jerasley@microsoft.com> Co-authored-by: Jeff Rasley <jerasley@microsoft.com> Co-authored-by: Shaden Smith <Shaden.Smith@microsoft.com> Co-authored-by: Jon Eyolfson <eyolfson@gmail.com> Co-authored-by: Stas Bekman <stas00@users.noreply.github.com> Co-authored-by: Cheng Li <pistasable@gmail.com> Co-authored-by: TheDudeFromCI <thedudefromci@gmail.com> Co-authored-by: Olatunji Ruwase <olruwase@microsoft.com> Co-authored-by: Sean Naren <sean@grid.ai>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adds an else statement to make sure we only call
init_distributedif required.I added this due to running into an
mpi4pymissing requirement issue, since we do not expose theauto_mpi_discoveryflag within theinit_distributedfunction. As a result by ensuring we use the check first to see if distributed has already been initialized, we can omit calling this check.