Evaluating large language models with Ray in hybrid cloud | Anyscale | Podwise