TT - Forge

TT-Forge is an open-source, MLIR-based compiler stack developed by Tenstorrent to streamline the compilation, optimization, debugging, and extension of machine learning models for custom silicon. It is highly modular, supports multiple AI frameworks such as PyTorch, TensorFlow, and ONNX, and is engineered for efficient, hardware-aware model deployment on Tenstorrent AI chips. Designed for developers, researchers, and AI/ML engineers, it provides tooling for both performance analysis and simulation, making it especially suitable for AI model execution and custom hardware innovation.