Pruning and Quantizing ML Models With One Shot Without Retraining

Discover advanced pruning and quantization techniques for machine learning models in this 53-minute video presentation by Neural Magic. Learn how to achieve significant model compression without retraining, using a one-shot approach that removes 60% of weights and quantizes the entire model to INT8 while maintaining 99% accuracy. Explore practical examples in Computer Vision and Natural Language Processing that demonstrate how these methods can lead to a 4X speedup in model performance. Gain valuable insights into implementing these time-efficient techniques, requiring only minutes of work, to enhance your current projects and research in the field of machine learning optimization.

Pruning and Quantizing ML Models With One Shot Without Retraining