Ecosyste.ms: Timeline
Browse the timeline of events for every public repo on GitHub. Data updated hourly from GH Archive.
frostedoyster pushed 1 commit to main metatensor/metatrain
- Print learning rate decreases (#365) 6b5fb29
frostedoyster closed an issue on metatensor/metatrain
Print learning rate decreases
We should log whenever the learning rate is decreased in our models to better diagnose potential problems during training runs from users.frostedoyster closed a pull request on metatensor/metatrain
Print learning rate decreases
Fixes #327 # Contributor (creator of pull-request) checklist - [ ] Tests updated (for new features and bugfixes)? - [x] Issue referenced (for PRs that solve an issue)? <!-- readthedoc...frostedoyster closed a pull request on metatensor/metatrain
Add MAE logging
Implements logging of MAEs. This is optional during training, and it defaults to False due to the already large amount of information being printed (training/validation loss/energy/force/stress etc...frostedoyster opened a pull request on metatensor/metatrain
Print learning rate decreases
Fixes #327 # Contributor (creator of pull-request) checklist - [ ] Tests updated (for new features and bugfixes)? - [x] Issue referenced (for PRs that solve an issue)?frostedoyster created a branch on metatensor/metatrain
print-lr - Training and evaluating machine learning models for atomistic systems.
frostedoyster opened an issue on metatensor/metatrain
A design for more generic targets
In order to extend the functionality of metatrain, it would be ideal to have infrastructure supporting more generic targets. In general, a target is defined by its metadata: - keys: for example...frostedoyster pushed 5 commits to nanopet metatensor/metatrain
frostedoyster pushed 1 commit to auto-continue metatensor/metatrain
- Make it distributed-proof 427d3d8
frostedoyster pushed 1 commit to auto-continue metatensor/metatrain
- Make it distributed-proof c33bcf2
frostedoyster pushed 1 commit to auto-continue metatensor/metatrain
- Call processing function before new `outputs/` directory is created 3c136bf
frostedoyster opened a draft pull request on metatensor/metatrain
Automatic continuation
Implements "automatic" continuation as described in #362. Questions: - is the naming good? - is the code good? - how and where should this be documented? # Contributor (creator of pull-r...frostedoyster created a branch on metatensor/metatrain
auto-continue - Training and evaluating machine learning models for atomistic systems.
frostedoyster opened an issue on metatensor/metatrain
"Automatic" continuation
Many packages implement automatic continuation, meaning that training is continued automatically if a checkpoint is detected. While I don't think this is a good default, it does allow for easy job ...abmazitov created a comment on a pull request on metatensor/metatrain
Thank you @frostedoyster! This looks good to me and I think we can merge without hesitation
frostedoyster pushed 3 commits to llpr metatensor/metatrain
frostedoyster pushed 1 commit to nanopet metatensor/metatrain
- Merge branch 'mae-logging' into nanopet 63bcfac
frostedoyster opened a pull request on metatensor/metatrain
Add MAE logging
Implements logging of MAEs. This is optional during training, and it defaults to False due to the already large amount of information being printed (training/validation loss/energy/force/stress etc...frostedoyster created a branch on metatensor/metatrain
mae-logging - Training and evaluating machine learning models for atomistic systems.