ai podcast by Kirill Solodskikh

Quantization is a powerful technique for reducing memory usage and speeding up AI applications built with LLMs, diffusion models, CNNs, and other architectures. In fact, quantization is fundamental to all data compression—from JPEG and GIF to MP3 and MP4 (HEVC)! In this episode, we'll cover the basics of neural network quantization, laying the groundwork for future episodes where we'll dive into specific quantization algorithms.