Statistics averaging #229
Merged
Statistics averaging #229
Conversation
This file contains 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
| @@ -46,7 +49,7 @@ def __init__(self, opt: torch.optim.Optimizer, *, average_parameters: bool, aver | |||
| def step(self, wait: bool = True, **kwargs): | |||
| """ Average optimizer weights and gradients with peers. """ | |||
| if not wait: | |||
| return run_in_background(self.step, wait=False, **kwargs) | |||
| return run_in_background(self.step, wait=True, **kwargs) | |||
justheuristic
approved these changes
Apr 17, 2021
justheuristic
requested review from
mryab and
borzunov
and removed request for
mryab
borzunov
reviewed
Apr 18, 2021
| grad_avg = 0.5 * (x1.grad + x2.grad) | ||
| stats_avg = 0.5 * (opt1.state[x1]["exp_avg_sq"] + opt2.state[x2]["exp_avg_sq"]) | ||
|
|
||
| f1 = averager1.step(wait=False) |
There was a problem hiding this comment.
Suggested change
| f1 = averager1.step(wait=False) | |
| # We set wait=False to test hivemind.utils.run_in_background() usage | |
| f1 = averager1.step(wait=False) |
nit: Using wait=False and then waiting for the result looked surprising to me. I'd suggest clarifying this with the comment.
There was a problem hiding this comment.
Actually the main purpose of wait=False is to prevent deadlock, when averager1 waits for averager2 to join.
Fix: write this explicitly in comment
There was a problem hiding this comment.
Oh, sure, missed that! Thanks for the explanation :)
borzunov
approved these changes
Apr 18, 2021
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This PR extends
TrainingAveragerwith optimizer's stats averaging. It should serve as base for subsequent decentralized adaptive optims implementations.