accelforge 2026.1.13.139

AccelForge

AccelForge

AccelForge is a framework to model and design tensor algebra accelerators. Key features include:

• Flexible and user-defined specifications for components, architectures, and workloads, including a suite of example accelerators and deep neural networks.
• An easy-to-use Python API for specifying and manipulating input specifications.
• Novel mapping algorithms that maps workloads onto architectures optimally, and in orders-of-magnitude less time than prior approaches.

To learn more, see the AccelForge documentation.

AccelForge uses HWComponents as a backend to model the area, energy, latency, and leak power of hardware components.

Installation

AccelForge is available on PyPI:

pip install accelforge

Notebooks and Examples

Examples can be found in the notebooks directory. Examples of the input files can be found in the examples directory.