Until recently, writing GPU kernels for LLM training and inference meant learning low-level programming tools like CUDA and C++. Triton, an open source, Python-based DSL created by OpenAI, bridges the gap between high-level machine learning frameworks and low-level GPU programming. Triton is built into PyTorch 2 and backed by NVIDIA, Intel, AMD, and Red Hat.
Harshwardhan Fartale
If you’re working with AI training or inference and want more control over GPU performance, you might be interested in GPU Programming with Triton by Harshwardhan Fartale.
The book explores how to use Triton to write efficient custom GPU kernels in Python—without having to work directly in CUDA C++. Readers will learn how GPUs execute programs, how to build and optimize Triton kernels, and how to apply those skills to real AI workloads. It also covers practical techniques for identifying bottlenecks and improving the performance of training and inference.
It’s currently available through Manning’s Early Access Program, so the content will continue to develop as new chapters are released.
- Full details: GPU Programming with Triton - Harshwardhan Fartale
Don’t forget you can get 45% off with your Devtalk discount! Just use the coupon code “devtalk.com” at checkout 