Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Duration 21 hours
Course Outline
Fundamentals of Cambricon and MLU Architecture
- An overview of the Cambricon AI chip portfolio
- Detailed look at MLU architecture and instruction pipelines
- Compatible model types and their ideal use cases
Setting Up the Development Toolchain
- Installation procedures for BANGPy and the Neuware SDK
- Configuring environments for both Python and C++
- Ensuring model compatibility and preprocessing strategies
Building Models with BANGPy
- Managing tensor structures and data shapes
- Constructing efficient computation graphs
- Implementing custom operations within BANGPy
Deployment via the Neuware Runtime
- Strategies for converting and loading models
- Controlling execution and inference processes
- Best practices for edge and datacentre deployment
Enhancing Performance
- Optimising memory mapping and layer tuning
- Utilising execution tracing and profiling tools
- Identifying common bottlenecks and applying effective fixes
Integrating MLUs into Applications
- Leveraging Neuware APIs for seamless application integration
- Supporting streaming and multi-model scenarios
- Managing hybrid CPU-MLU inference workloads
End-to-End Project and Practical Use Case
- Hands-on lab: Deploying a vision or NLP model
- Implementing edge inference with BANGPy integration
- Evaluating both accuracy and throughput
Recap and Path Forward
Requirements
- A solid grasp of machine learning model architectures
- Practical experience with Python and/or C++
- Knowledge of model deployment and acceleration principles
Target Audience
- Embedded AI developers
- ML engineers focused on edge or datacentre deployment
- Developers working with Chinese AI infrastructure
Testimonials (1)
That we can cover advance topic and work with real-life example
