TT-Metalium™ is Tenstorrent's open-source, low-level AI hardware SDK designed to provide developers with direct access to the underlying hardware components for custom kernel development and experimentation. Unlike other platforms, TT-Metalium™ offers complete transparency with no black boxes, encrypted APIs, or hidden functions, making it ideal for developers aiming to customize models, create new ones, or run non-machine learning code.
TT-Forge™ is Tenstorrent's MLIR-based compiler, designed to seamlessly integrate with various machine learning (ML) frameworks, including PyTorch, JAX, TensorFlow, and ONNX. It serves as a bridge between high-level models and Tenstorrent hardware, facilitating efficient compilation and execution of AI workloads. By leveraging open-source technologies such as OpenXLA, LLVM’s MLIR, and torch-mlir, TT-Forge™ offers a flexible and extensible foundation for AI development. Its architecture supports the ingestion of models from multiple frameworks, enabling a structured approach to compilation and optimization. This integration ensures that developers can build and deploy AI models with ease, maximizing performance on Tenstorrent hardware.
Tenstorrent is a technology company specializing in advanced computing solutions, particularly in the realm of artificial intelligence and machine learning. The company focuses on developing high-performance hardware and software architectures designed to accelerate AI workloads. Tenstorrent aims to provide innovative solutions that enhance computational efficiency and scalability for various industries, including data centers and enterprise applications.