RC RANDOM CHAOS

Raschka Releases 11-Hour Reasoning Model Video Course

· via Hacker News

Original source

Sebastian Raschka's from-scratch video course on building reasoning models

Hacker News →

Sebastian Raschka released an eight-part video course totaling 11 hours on implementing reasoning models from scratch using PyTorch. Reasoning models enhance standard language models with reinforcement learning and inference scaling—capabilities now standard in modern LLMs. The course explores how these models function as the engine for AI agent systems, which layer additional tools and context management around them.

The series starts with a pre-trained Qwen3 0.6B model rather than training from scratch, allowing developers to focus on reasoning techniques like inference-time scaling, reinforcement learning, and distillation. Topics include text generation, key-value caching, mathematical verification, and evaluation. The course assumes familiarity with Python and PyTorch; a corresponding Manning book provides deeper explanations and exercises, while a shorter LinkedIn Learning course offers a 90-minute conceptual overview for those seeking faster context.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.