Metadata-Version: 2.4
Name: runnerase-training
Version: 0.8.0
Summary: Training calculators for the Ruhr University Neural Network energy representation (RuNNer).
Author: Alexander L. M. Knoll
Author-email: Alexander L. M. Knoll <alexander.knoll@rub.de>
License-Expression: GPL-3.0-or-later
License-File: LICENSE
Classifier: Topic :: Scientific/Engineering :: Chemistry
Classifier: Intended Audience :: Science/Research
Classifier: Development Status :: 4 - Beta
Classifier: Operating System :: POSIX
Classifier: Operating System :: MacOS
Classifier: Programming Language :: Python :: 3
Requires-Dist: ase>=3.29.0
Requires-Dist: numpy>=2.0.0
Requires-Dist: runnerase-core
Requires-Python: >=3.10
Project-URL: Bugs, https://gitlab.com/runner-suite/runnerase-2.0/issues
Project-URL: Documentation, https://runner.theochem2.rub.de/runnerase-docs/latest/
Project-URL: Source, https://gitlab.com/runner-suite/runnerase-2.0
Description-Content-Type: text/markdown

# runnerase-training

File-based calculator and analysis tools for training and evaluating RuNNer
neural network potentials, built on top of `runnerase-core`. While
`runnerase-prediction` evaluates an already-trained potential in-process,
`runnerase-training` drives RuNNer's training binary directly and helps you
inspect what came out of a fit.

## Provides

- `RuNNerbin`: a file-based calculator that manages running RuNNer's training
  binary and the corresponding input/output files.
- Storage classes for analyzing RuNNer output: `FitResults`, `ResultsClass`,
  `ScalingData`, `QEqScalingData`, `DatasetSplit`,
  `RuNNerSymmetryFunctionValues`, `Weights`, `ElementalWeights`.

## Usage

```python
from runnerase_training import RuNNerbin

calc = RuNNerbin(directory="training_run")
calc.calculate()  # runs the RuNNer.x training binary

results = calc.results  # FitResults, weights, scaling data, ...
```

## Documentation

Tutorials on training and workflow management with `RuNNerbin`:
https://runner.theochem2.rub.de/runnerase-docs/latest/

## License

GPL-3.0-or-later. Part of the [runnerase](https://gitlab.com/runner-suite/runnerase-2.0) project.
