pseudodynamics._config

Classes

ExperimentConfig([config, args, model])

Records and manages experiment configuration settings.
class pseudodynamics._config.ExperimentConfig(config=None, args=None, model=None)[source]

Bases: object

Records and manages experiment configuration settings.

Parameters:
find_lastest_ckpt()[source]

Find the ckpt file with the minimum val_loss.

Fallback: when val_loss was never logged (e.g. cord-blood PCA whose val_dataloader is empty), the ModelCheckpoint monitor fires no match and only last.ckpt exists (saved via save_last=True). In that case we return last.ckpt so eval does not fail with FileNotFoundError.

from_json(file_path, main_dir='/home/docs/checkouts/readthedocs.org/user_builds/pseudodynamics-plus/envs/latest/lib/python3.13/')[source]

Load a saved experiment configuration from JSON file.

Return type:

ExperimentConfig

Parameters:

  • file_path (str)

  • main_dir (str)

Args:

file_path: Path to the saved JSON configuration file

Returns:

ExperimentConfig instance with loaded parameters

save(path)[source]

Saves configuration to JSON file.

Parameters:

path (str)

to_dict()[source]

Returns all configurations as a dictionary.

Return type:

Dict[str, Any]