Skip to main content

Python package

max

The MAX Python API reference.

The MAX API provides a high-performance graph compiler and runtime library that executes AI models with incredible speed on a wide range of hardware.

MAX offers a layered architecture that lets you work at the level of abstraction that best fits your needs. From deploying production-ready models with a few lines of code to building custom neural networks from scratch, each layer builds upon the others so you can move between levels seamlessly as requirements evolve.

To install the MAX Python API, see the Quickstart guide.

Modules

  • driver: Low-level device management and tensor operations.
  • dtype: Unified data type system supporting various numeric formats.
  • engine: Model execution runtime with automatic optimization.
  • entrypoints: Command-line tools and serving infrastructure.
  • interfaces: Universal interfaces for consistent API integration.
  • torch: PyTorch integration for custom operations and interoperability.

Packages

  • graph: Computational graph construction with 100+ operations for complete model control.
  • pipelines: Pre-built, optimized model architectures for immediate deployment.
  • nn: High-level neural network building blocks with automatic graph compilation.