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 14 hours
Course Outline
Intro to Edge AI and the Ascend 310
- Edge AI landscape: trends, limitations, and use cases
- Architecture of the Huawei Ascend 310 chip and its compatible toolchain
- The role of CANN in the edge AI deployment ecosystem
Model Prep and Transformation
- Extracting trained models from TensorFlow, PyTorch, and MindSpore
- Converting models to OM format for Ascend devices using ATC
- Addressing unsupported operations and employing lightweight conversion methods
Creating Inference Pipelines with AscendCL
- Executing OM models on Ascend 310 via the AscendCL API
- Pre/post-processing I/O, memory management, and device control
- Deployment within embedded containers or minimal runtime environments
Refining for Edge Limitations
- Shrinking model size and adjusting precision (FP16, INT8)
- Leveraging the CANN profiler to pinpoint performance bottlenecks
- Optimizing memory layout and data streaming for efficiency
Deployment via MindSpore Lite
- Utilizing the MindSpore Lite runtime for mobile and embedded platforms
- Evaluating MindSpore Lite against the raw AscendCL pipeline
- Packaging inference models for target-specific deployment
Edge Deployment Scenarios and Practical Examples
- Case study: Object detection model on a smart camera with Ascend 310
- Case study: Real-time classification within an IoT sensor hub
- Overseeing and updating models deployed at the edge
Recap and Future Directions
Requirements
- Background in AI model development or deployment processes
- Fundamental understanding of embedded systems, Linux, and Python
- Experience with deep learning frameworks like TensorFlow or PyTorch
Target Audience
- IoT solution developers
- Embedded AI engineers
- Edge system integrators and AI deployment specialists
Testimonials (1)
That we can cover advance topic and work with real-life example
