Ecosystem
Move from theory to implementation through layers that work together or independently.
P10
The foundational tensor engine for memory, operators, dispatch, and hardware backends.
TPX
The explicit automatic differentiation layer for dynamic graphs.
Stax
The static graph, JIT, operator fusion, and capture playground.
NN
The high-level neural network API for building models while keeping the underlying computation traceable.
Learning to hardware
Pythonic APIs, readable CPU and CUDA backends, and nanobind integration connect fundamentals, custom layers, and hardware kernels.