Add interpretability to deep learning script

I have a keras script that analyzes spectrogram from ECG signal. I need an interpretability module with lime, shap and tf-explain or whatever you propose. Also i need an export of the model prediction for the testing dataset.