Feature/learner updates #184
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
patbaa
commented
Dec 8, 2021
| Recompiles the Keras model. This way the optimizer history get erased, | ||
| which is needed before a new training round, otherwise the outdated history is used. | ||
| """ | ||
| compile_args = self.model._get_compile_args() # pylint: disable=protected-access |
There was a problem hiding this comment.
I did not find any better way of resetting the optimizer, than re-compiling the mode, which needs calling this protected function.
When I simply replace the optimizer
new_opt = keras.optimizers.SGD(lr=0.1)
model.optimizer = new_optthen I cannot change batch size anymore, and also the optimizer seems not to be updated, the model.optimizer.get_weights() should store the number of steps and the momentums, but it remains an empty list this way.
There was a problem hiding this comment.
maybe name the function reset_optimizer to better reflect why we have this and when we want to use it
qati
reviewed
Dec 10, 2021
| Recompiles the Keras model. This way the optimizer history get erased, | ||
| which is needed before a new training round, otherwise the outdated history is used. | ||
| """ | ||
| compile_args = self.model._get_compile_args() # pylint: disable=protected-access |
There was a problem hiding this comment.
maybe name the function reset_optimizer to better reflect why we have this and when we want to use it
evsmithx
approved these changes
Dec 15, 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Outdated suggestions cannot be applied.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.
Two changes:
1.) The
mli_get_current_weightsreturns all trainable and non-trainable parameters. Previously only the trainable parameters were returned for PyTochLearner, but normalization layers have non-trainable parameters, which are needed too to reproduce the model performance. See running statistics here.2.) Erase optimizer history at the beginning of the
mli_propose_weights, which is more general, as history gets outdated when others improve the model.The learner test files were also modified, accordingly.