Cambricon MLU Development with BANGPy and Neuware

• An understanding of machine learning model structures
• Experience with Python and/or C++
• Familiarity with model deployment and acceleration concepts

Audience

• Embedded AI developers
• ML engineers deploying to edge or datacenter
• Developers working with Chinese AI infrastructure

Cambricon MLUs (Machine Learning Units) are specialized AI chips optimized for inference and training in edge and datacenter scenarios.

This instructor-led, live training (online or onsite) is aimed at intermediate-level developers who wish to build and deploy AI models using the BANGPy framework and Neuware SDK on Cambricon MLU hardware.

By the end of this training, participants will be able to:

• Set up and configure the BANGPy and Neuware development environments.
• Develop and optimize Python- and C++-based models for Cambricon MLUs.
• Deploy models to edge and data center devices running Neuware runtime.
• Integrate ML workflows with MLU-specific acceleration features.

Format of the Course

• Interactive lecture and discussion.
• Hands-on use of BANGPy and Neuware for development and deployment.
• Guided exercises focused on optimization, integration, and testing.

Course Customization Options

• To request a customized training for this course based on your Cambricon device model or use case, please contact us to arrange.

Introduction to Cambricon and MLU Architecture

• Overview of Cambricon’s AI chip portfolio
• MLU architecture and instruction pipeline
• Supported model types and use cases

Installing the Development Toolchain

• Installing BANGPy and Neuware SDK
• Environment setup for Python and C++
• Model compatibility and preprocessing

Model Development with BANGPy

• Tensor structure and shape management
• Computation graph construction
• Custom operation support in BANGPy

Deploying with Neuware Runtime

• Converting and loading models
• Execution and inference control
• Edge and data center deployment practices

Performance Optimization

• Memory mapping and layer tuning
• Execution tracing and profiling
• Common bottlenecks and fixes

Integrating MLU into Applications

• Using Neuware APIs for application integration
• Streaming and multi-model support
• Hybrid CPU-MLU inference scenarios

End-to-End Project and Use Case

• Lab: Deploying a vision or NLP model
• Edge inference with BANGPy integration
• Testing accuracy and throughput

Summary and Next Steps