NeuroWeave SDK

The NeuroWeave SDK is an integral component of the Cadence AI IP platform, designed to facilitate the development and deployment of AI networks across various Cadence AI IP offerings. It provides a comprehensive suite of tools, libraries, and drivers that enable efficient mapping of machine learning networks to the available IP options, ensuring optimal performance, power efficiency, and area utilization.

This SDK supports popular AI frameworks such as PyTorch, TensorFlow, TensorFlow Light Micro (TFLM), and ONNX, allowing users to develop AI networks with familiar tools. Once developed, these networks can be quantized and mapped to the targeted Cadence AI IP, providing flexibility in choosing the best IP configuration for specific performance and power requirements.

The NeuroWeave SDK is designed to adapt to changing requirements, offering a straightforward path to explore different AI IP configurations. This adaptability ensures that users can easily adjust their AI solutions to meet evolving demands.

Key Features and Benefits

• Efficient mapping of state-of-the-art AI/ML workloads with high inferences per second and low latency.
• Industry-leading performance and power efficiency, measured in inferences per second per area (IPS/mm²) and per power (IPS/W).
• End-to-end software toolchain supporting a wide range of markets and frameworks.

The SDK is part of a broader AI IP platform that includes Neo NPUs, NeuroEdge AI Co-processor, and Tensilica DSPs, each offering unique capabilities to meet diverse application needs. The Neo NPU provides scalable AI acceleration, the NeuroEdge AI Co-processor handles complex operations not suited for NPUs, and Tensilica DSPs offer flexible instruction sets for mixed AI and DSP workloads.

Overall, the NeuroWeave SDK empowers developers to integrate AI into their systems efficiently, unlocking the full potential of AI integration across various applications and markets.