New active learning with AiiDA and FLARE#414
Merged
Merged
Conversation
I implement in the `AiiDAEnsemble` a first version of on-the-fly training using the FLARE code, which exploits (sparse) Gaussian processes and the ACE/MB descriptors to train and make estimations of the errors.
…sscha into new/flare-interface
Examples for running SSCHA using AiiDA+FLARE to perform the SSCHA minimization are added. These examples serve as well as for testing some hard-to-test logic within the OTF workflow.
The mir-flare python package must be installed to perform the tests during the continous integration workflow.
…to new/flare-interface
…to new/flare-interface
In AiiDAEnsemble it is added the possibility of splitting the submission of all the calculations in a certain number of batches. This is convenient especially when performing the calculations using active-learning/on-the-fly simulations.
The `self.init()` wasn't called, and it was messing up the minimization.
A variable controlling when the training is performed is added. This can give improvements if the hyperparameter training gets stucked into local minima due to few environments in the dataset.
…to new/flare-interface
…to new/flare-interface
mesonepigreco
self-requested a review
April 18, 2026 19:01
Collaborator
|
@bastonero The code is not passing some of the tests. Can you check so that we can merge in the official one before release 1.7? |
Contributor
Author
|
@mesonepigreco The blocking issue is that the FLARE code doesn't install successfully. I now changed the installation to build it directly from the cloned repo (which should work). Hope this works |
Contributor
Author
|
@mesonepigreco ci siamo! |
mesonepigreco
approved these changes
Jul 20, 2026
mesonepigreco
left a comment
mesonepigreco
left a comment
Collaborator
There was a problem hiding this comment.
It seems everything good. Ready to merge
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
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.
We here implement a new active learning protocol using the FLARE machine-learning interatomic potential. FLARE provides prediction uncertainties that can be used to actively learn the potential energy surface of the material.
The description of this new scheme will be made available soon through an arXiv publication. "Lorenzo Bastonero, Gabriel Joalland, Chiara Cignarella, Lorenzo Monacelli, and Nicola Marzari, Predicting challenging phase transitions in energy materials with Bayesian active learning".
Examples are provided within the dedicated folder.