XQuant Report Keras¶

model_compression_toolkit.xquant.keras.facade_xquant_report.xquant_report_keras_experimental(float_model, quantized_model, repr_dataset, validation_dataset, xquant_config)¶

Generate an explainable quantization report for a quantized Keras model.

Parameters:

• float_model (keras.Model) – The original floating-point Keras model.

• quantized_model (keras.Model) – The quantized Keras model.

• repr_dataset (Callable) – The representative dataset used during quantization for similarity metrics computation.

• validation_dataset (Callable) – The validation dataset used for evaluation for similarity metrics computation.

• xquant_config (XQuantConfig) – Configuration settings for explainable quantization.

Returns:

A dictionary containing the collected similarity metrics and report data.

Return type:

Dict[str, Any]

Return type:

Dict[str, Any]