pyreal.RealApp#

class pyreal.RealApp(models, X_train_orig=None, y_train=None, transformers=None, feature_descriptions=None, active_model_id=None, classes=None, class_descriptions=None, pred_format_func=None, fit_transformers=False, id_column=None, openai_api_key=None, llm=None, context_description='')[source]#

Maintains all information about a Pyreal application to generate explanations

__init__(models, X_train_orig=None, y_train=None, transformers=None, feature_descriptions=None, active_model_id=None, classes=None, class_descriptions=None, pred_format_func=None, fit_transformers=False, id_column=None, openai_api_key=None, llm=None, context_description='')[source]#

Initialize a RealApp object

Parameters:

  • object (models (model) – model): Model(s) for this application

  • models (list of) – model): Model(s) for this application

  • model_id (or dict of) – model): Model(s) for this application

  • X_train_orig (DataFrame of shape (n_instances,n_features) – Training data for models. If None, must be provided when preparing explainers.

  • y_train (DataFrame of shape (n_instances,)) – The y values for the dataset

  • transformers (Transformer object or list of Transformer objects) – Transformers for this application

  • feature_name (feature_descriptions (dictionary of) – feature_description): Mapping of default feature names to readable names

  • active_model_id (string or int) – ID of model to store as active model, if None, this is set to the first model

  • classes (array) – List of class names returned by the model, in the order that the internal model considers them if applicable. Can be automatically extracted if model is an sklearn classifier None if model is not a classifier

  • class_descriptions (dict) – Interpretable descriptions of each class None if model is not a classifier

  • pred_format_func (function) – Function to format model prediction outputs

  • fit_transformers (Boolean) – If True, fit the transformers to X_train_orig on initialization

  • id_column (string or int) – Name of column that contains item ids in input data

  • openai_api_key (string) – OpenAI API key. Required for GPT narrative explanations, unless openai client is provided

  • llm (LLM model object) – Local LLM object or LLM client object to use to generate narratives.

  • context_description (string) – Description of the model’s prediction task, in sentence format. This is used by LLM model for narrative explanations. For example: “The model predicts the price of houses.”

Methods

__init__(models[, X_train_orig, y_train, ...])

Initialize a RealApp object

add_model(model[, model_id])

Add a model

from_sklearn([pipeline, model, ...])

Create a RealApp from a sklearn pipeline or model and transformers. Must provide one of: - just pipeline - just model - model and transformers.

get_active_model()

Return the active model

predict(x[, model_id, as_dict, format])

Predict on x using the active model or model specified by model_id

predict_proba(x[, model_id, as_dict, format])

Return the predicted probabilities of x using the active model or model specified by model_id, only if the model has a predict_proba method

prepare_feature_contributions([...])

Initialize and fit a local feature contribution explainer

prepare_feature_importance([x_train_orig, ...])

Initialize and fit a global feature importance explainer

prepare_similar_examples([x_train_orig, ...])

Initialize and fit a nearest neighbor explainer

produce_feature_contributions(x_orig[, ...])

Produce a feature contribution explanation

produce_feature_importance([model_id, ...])

Produce a GlobalFeatureImportance explainer

produce_narrative_feature_contributions(x_orig)

Produce a feature contribution explanation, formatted in natural language sentence format using LLMs.

produce_similar_examples(x_orig[, model_id, ...])

Produce a SimilarExamples explainer

set_active_model_id(active_model_id)

Set a new active model

set_openai_client([openai_client, ...])

Set the openai client for this RealApp.

train_feature_contribution_llm([...])

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