srush/GPU-Puzzles
An interactive Jupyter notebook tutorial that teaches GPU programming through puzzles, using Numba to map Python to CUDA kernels.

Velocity · 7d
+8.6
★ / day
Trend
→steady
star history
The repository contains a series of hands-on coding exercises that teach GPU programming from basics to advanced concepts. Rather than exposition, it immerses learners in writing CUDA kernels directly using Numba, which compiles Python to GPU code. The puzzles progressively build intuition for the GPU architectures that underpin modern deep learning systems.