"""Utility functions for perturbation based predictions."""
import numpy as np
from tqdm import tqdm
[docs]
def make_predictions(data, runner, batch_size):
"""Make predictions with the input data.
Process the data with the model runner and return the predictions.
Args:
data (np.ndarray): An array of masked input data to be processed by the model.
runner (object): An object that runs the model on the input data and returns predictions.
batch_size (int): The number of masked inputs to process in each batch.
Returns:
np.ndarray: An array of predictions made by the model on the input data.
"""
number_of_masks = len(data)
predictions = []
for i in tqdm(range(0, number_of_masks, batch_size), desc='Explaining'):
predictions.append(runner(data[i:i + batch_size]))
return np.concatenate(predictions)