Watch this video on YouTube
What should be considered when using SparkTrials on GPU-enabled clusters, and how should parallelism be configured?