Hydro - Surrogate-Based Hyperparameter Tuning Service in Datacenters
Offered By: USENIX via YouTube
Course Description
Overview
Explore a conference talk from OSDI '23 that introduces Hydro, an innovative surrogate-based hyperparameter tuning service designed for datacenters. Learn about the challenges of hyperparameter tuning for large-scale deep learning models and how Hydro addresses these issues through job-level and cluster-level optimizations. Discover the two key components of Hydro: the Tuner, which generates and optimizes surrogate models, and the Coordinator, which improves tuning efficiency and resource utilization. Gain insights into the experimental results that demonstrate Hydro's ability to significantly reduce tuning makespan without compromising quality. Access the source code and delve into the potential impact of this technology on deep learning model development and datacenter resource management.
Syllabus
OSDI '23 - Hydro: Surrogate-Based Hyperparameter Tuning Service in Datacenters
Taught by
USENIX
Related Courses
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and OptimizationDeepLearning.AI via Coursera Machine Learning in the Enterprise
Google Cloud via Coursera Art and Science of Machine Learning 日本語版
Google Cloud via Coursera Art and Science of Machine Learning auf Deutsch
Google Cloud via Coursera Art and Science of Machine Learning en Español
Google Cloud via Coursera