This repository has been archived by the owner on Nov 3, 2023. It is now read-only.
Fix LR scheduler cooldown #3719
Merged
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
emilydinan
approved these changes
Jun 15, 2021
klshuster
approved these changes
Jun 15, 2021
| if optim_states and saved_optim_type != opt['optimizer']: | ||
| # we changed from adam to adamax, or sgd to adam, or similar | ||
| logging.warning('Not loading optim state since optim class changed.') | ||
| return False | ||
| return True | ||
| elif optim_states: | ||
| # check for any fp16/fp32 conversions we need to do | ||
| optimstate_fp16 = 'loss_scaler' in optim_states |
There was a problem hiding this comment.
i can't leave a comment for it, but are the semantics correct for line 1099? the elif not optimstate_fp16 and self.fp16 block? are we returning True always because of the lower precision conversion?
| self.scheduler = optim.lr_scheduler.LambdaLR(optimizer, self._linear_lr) | ||
|
|
||
| def _linear_lr(self, step): | ||
| # this multiplicative factor ensures linear decay rate | ||
| # lr_mult = float(self.max_lr_steps - step - 1) / float(self.max_lr_steps - step) | ||
| lr_mult = max(0.0, 1e-6 + (1.0 - step / self.max_lr_steps) * (1 - 1e-6)) | ||
| lr_mult = max(0.0, 1.0 - step / self.max_lr_steps) |
There was a problem hiding this comment.
we dont need 1e-6 anymore?
There was a problem hiding this comment.
I made an executive call to let it actually go to 0 :P
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
Patch description
Context:
Testing steps
Adjusted CI, new assertions.