Boosting machine learning workflows with GPU-accelerated libraries