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.
Course Outline
Introduction to Cambricon and MLU Architecture
- Overview of Cambricon's AI chip lineup.
- Deep dive into MLU architecture and instruction pipelines.
- Supported model types and their respective use cases.
Setting Up the Development Toolchain
- Installation of BANGPy and the Neuware SDK.
- Configuring development environments for Python and C++.
- Model compatibility checks and preprocessing workflows.
Model Development with BANGPy
- Managing tensor structures and shapes.
- Constructing computation graphs.
- Implementing custom operations within BANGPy.
Deployment via Neuware Runtime
- Model conversion and loading procedures.
- Controlling execution and inference.
- Best practices for edge and data center deployment.
Performance Optimization
- Optimizing memory mapping and layer tuning.
- Utilizing execution tracing and profiling tools.
- Identifying common bottlenecks and applying solutions.
Integrating MLU into Applications
- Leveraging Neuware APIs for seamless application integration.
- Supporting streaming and multi-model architectures.
- Designing hybrid CPU-MLU inference scenarios.
End-to-End Project and Use Case
- Lab session: Deploying a vision or NLP model.
- Implementing edge inference with BANGPy integration.
- Testing and validating accuracy and throughput.
Summary and Next Steps
Requirements
- A solid understanding of machine learning model architectures.
- Proficiency in Python and/or C++.
- Familiarity with concepts related to model deployment and acceleration.
Audience
- Embedded AI developers.
- ML engineers focused on edge or data center deployments.
- Developers working within Chinese AI infrastructure ecosystems.
21 Hours
Testimonials (1)
That we can cover advance topic and work with real-life example
