NeuralFlix
Intro to Deep Learning Model Sparsification
Presenter: Mark Kurtz
Learn about deep learning model sparsification and how it helps you deliver SOTA inference performance on commodity CPUs. Mark gives an overview of sparsification and its different techniques, including pruning, quantization, and knowledge distillation. He also introduces an approach called "compound sparsification" that combines all techniques together for best-in-class results.