Input files

XANESNET uses YAML configuration files for three workflow modes: training, inference, and analysis. Each mode has its own top-level sections.

Every configuration is validated against the schemas in xanesnet/schemas/. The schemas define allowed fields, types, and default values.

The required top-level sections are: * training: datasource, dataset, model, trainer, strategy, and encodings * inference: datasource, dataset, and inferencer * analysis: selectors, collectors, aggregators, reporters, and plotters

Many nested fields have defaults, so they do not need to be written when the default is suitable. Example files are available in configs/. The command-line workflows that consume these files are described in Running XANESNET, and schema validation is implemented by validate_config_schema().

Configuration sections

The detailed configuration pages are organised by top-level section:

Section	Purpose	Details
datasource	Locate raw structures and spectra.	Datasources
dataset	Prepare, represent, cache, and split the data.	Datasets
encodings	Transform spectra before model input or loss calculation.	Encodings
model	Select the neural network and its parameters.	Models
trainer	Configure optimisation, losses, validation, and stopping.	Trainer
strategy	Select single-model or ensemble training behaviour.	Training strategies
inferencer	Configure prediction from a trained checkpoint.	Inferencer
Analysis lists	Select, score, summarise, report, and plot predictions.	Analysis configuration