How to Make Your Model Fast
A free, open-source book on machine learning performance engineering that systematically introduces ML performance optimization methodologies spanning from hardware fundamentals to agent applications. Topics include Roofline analysis, quantization, pruning, kernel optimization, and serving deployment.
Timeline (last 90 days)
The free open-source book How to Make Your Model Fast: A Systems View of Efficient Machine Learning, from Silicon to Agents is hosted on GitHub (usamahz/make-your-model-fast) and is completely free and open source.
The book distinguishes model performance bottlenecks into four types: compute-bound, bandwidth-bound, memory-bound, and system-bound.
The book adopts a bottom-up structure, covering topics including kernels, compilers, quantization and pruning, vision models, on-device LLMs, robotics, performance profiling, serving deployment, and agents.
The author publicly solicits feedback and contributions, particularly welcoming readers working in ML systems, inference, compilers, edge AI, or performance engineering to participate in co-building.
All Facts (4)
The book distinguishes model performance bottlenecks into four types: compute-bound, bandwidth-bound, memory-bound, and system-bound.
50%UnverifiedThe book adopts a bottom-up structure, covering topics including kernels, compilers, quantization and pruning, vision models, on-device LLMs, robotics, performance profiling, serving deployment, and agents.
50%UnverifiedThe author publicly solicits feedback and contributions, particularly welcoming readers working in ML systems, inference, compilers, edge AI, or performance engineering to participate in co-building.
50%UnverifiedThe free open-source book How to Make Your Model Fast: A Systems View of Efficient Machine Learning, from Silicon to Agents is hosted on GitHub (usamahz/make-your-model-fast) and is completely free and open source.
50%