Models¶
- class deepdiagnostics.models.model.Model(model_path)¶
Load a pre-trained model for analysis.
- Parameters:
model_path (str) – relative path to a model.
- class deepdiagnostics.models.SBIModel(model_path)¶
Load a trained model that was generated with Mackelab SBI [TCBD+20]. Read more about saving and loading requirements here.
- Parameters:
model_path (str) – Relative path to a model - must be a .pkl file.
- predict_posterior(data, context_samples)¶
Sample the posterior and then
- Parameters:
data (deepdiagnostics.data.Data) – Data module with the loaded simulation
context_samples (np.ndarray) – X values to test the posterior over.
- Returns:
Simulator output
- Return type:
np.ndarray
- sample_posterior(n_samples, x_true)¶
Sample the posterior
- Parameters:
n_samples (int) – Number of samples to draw
x_true (np.ndarray) – Context samples. (must be dims=(n_samples, M))
- Returns:
Posterior samples
- Return type:
np.ndarray
- static save_posterior(neural_posterior, path, allow_overwrite=False)¶
Save an SBI posterior to a pickle file.
- Parameters:
neural_posterior (NeuralPosterior) – A neural posterior object. Must be an instance of the base class ‘NeuralPosterior’ from the sbi package.
path (str) – Relative path to a model - must be a .pkl file.
allow_overwrite (bool, optional) – Controls whether an attempt to overwrite succeeds or results in an error. Defaults to False.
- Return type:
None