How Well Do Sparse Models Transfer? - Exploring Transfer Performance in Computer Vision and NLP
Neural Magic via YouTube
Overview
Explore the transfer performance of sparse models in this 50-minute webinar recording from Neural Magic. Dive into the findings from two accepted papers: "How Well Do Sparse Imagenet Models Transfer" (CVPR 2022) and "The Optimal BERT Surgeon" (ENLP 2022). Discover how sparse models can match or surpass dense models in transfer performance, even at high sparsities, while offering significant inference and training speedups. Learn about the potential of sparse models in various applications and their implications for efficient machine learning. Engage with the Neural Magic team and community members through the Deep Sparse Community Slack for further discussions and questions.
Syllabus
How Well Do Sparse Models Transfer?
Taught by
Neural Magic