neural_compressor.experimental.nas.dynast.dynas_predictor

DyNAS Manager class.

Module Contents

Classes

Predictor

The Predictor class.

class neural_compressor.experimental.nas.dynast.dynas_predictor.Predictor(alphas=DEFAULT_ALPHAS, cost_factors=DEFAULT_COST_FACTORS, max_iterations=DEFAULT_MAX_ITERATIONS, verbose=False)

The Predictor class.

It handles the prediction of the metrics like accuracy, MACs and latency etc..

train(examples, labels)

Train the predictor on the specified examples and labels using the underlying regressor.

Parameters:

• examples – Examples to be used for training.

• labels – Labels to be used for training.

predict(examples)

Predict the output values of the specified examples using the underlying regressor.

Parameters:

examples – Examples for which predictions will be made.

Returns:

Predictions of the specified examples.

get_parameters()

Return the optimal parameter values of the underlying regressor.

Returns:

Optimal parameter values of the underlying regressor.

get_metrics(examples, labels)

Compute the performance metrics of the underlying regressor.

Parameters:

• examples – Examples to use when computing performance metrics.

• labels – Labels to use when computing performance metrics.

Returns:

Performance metrics of the underlying regressor. The metrics are

Mean absolute percentage error (MAPE) Root mean squared error (RMSE) Kendall rank correlation coefficient (kendall) Spearman’s rank correlation coefficient (spearman)

load(filename)

Load the model of the underlying regressor and searcher.

Parameters:

filename – Name of the file from which to load the model.

save(filename)

Save the model of the underlying regressor and searcher.

Parameters:

filename – Name of the file to which to save the model.