wf_psf.psf_models.psf_model_loader

PSF Model Loader.

This module provides helper functions for loading trained PSF models. It includes utilities to: - Load a model from disk using its configuration and weights. - Prepare inputs for inference or evaluation workflows.

Author: Jennifer Pollack <jennifer.pollack@cea.fr>

Functions

load_trained_psf_model(training_conf, ...)

Loads a trained PSF model and applies saved weights.
wf_psf.psf_models.psf_model_loader.load_trained_psf_model(training_conf, data_conf, weights_path_pattern)[source]

Loads a trained PSF model and applies saved weights.

Parameters:

  • training_conf (RecursiveNamespace) – Configuration object containing model parameters and training hyperparameters. Supports attribute-style access to nested fields.

  • data_conf (RecursiveNamespace or dict) – Configuration RecursiveNamespace object or a dictionary containing data parameters (e.g. pixel data, positions, masks, etc).

  • weights_path_pattern (str) – Glob-style pattern used to locate the model weights file.

Returns:

model – The PSF model instance with loaded weights.

Return type:

tf.keras.Model or compatible

Raises:

RuntimeError – If loading the model weights fails for any reason.