pyreal.RealApp.train_feature_contribution_llm#

RealApp.train_feature_contribution_llm(x_train_orig=None, live=True, provide_examples=False, num_inputs=5, num_features=3)[source]#

Run the training process for the LLM model used to generate narrative feature contribution explanations.

Parameters:

  • x_train_orig (DataFrame of shape (n_instances, n_features)) – Training set to take sample inputs from. If None, the training set must be provided to the explainer at initialization.

  • live (Boolean) – If True, run the training process through CLI input/outputs. If False, this function will generate a shell training file that will need to be filled out and added to the RealApp manually. Currently only live training is supported.

  • provide_examples (Boolean) – If True, generate a base example of explanations at each step. This may make the process faster, but will incur costs to your OpenAI API account.

  • num_inputs (int) – Number of inputs to request.

  • num_features (int) – Number of features to include per explanation. If None, all features will be included

Returns:

list of (explanation, narrative) pairs

The generated training data