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
Foundational Concepts of CANN and Ascend AI Processors
- Defining CANN and its position within Huawei’s AI computing infrastructure
- Introduction to Ascend processor architectures (e.g., 310, 910)
- Overview of supported AI frameworks and associated toolchains
Model Transformation and Compilation Processes
- Utilizing the ATC tool to convert models from TensorFlow, PyTorch, and ONNX
- Generation and verification of OM model files
- Addressing unsupported operators and resolving frequent conversion challenges
Integration with MindSpore and Compatible Frameworks
- Deploying models using MindSpore Lite
- Incorporating OM models via Python APIs or C++ SDKs
- Utilizing the Ascend Model Manager for workflow management
Enhancing Performance and Profiling Execution
- Exploring AI Core, memory management, and tiling optimization techniques
- Analyzing model execution through CANN profiling utilities
- Applying best practices to boost inference speed and minimize resource consumption
Managing Errors and Troubleshooting
- Identifying and fixing typical deployment errors
- Interpreting logs and employing error diagnostic tools
- Conducting unit tests and functional validation of deployed models
Strategies for Edge and Cloud Deployment
- Implementing solutions on Ascend 310 for edge computing applications
- Connecting with cloud-based APIs and microservices architectures
- Examining practical case studies in computer vision and NLP
Recap and Future Development Pathways
Requirements
- Proficiency with Python-based deep learning ecosystems like TensorFlow or PyTorch
- Solid grasp of neural network structures and model training processes
- Fundamental command of Linux CLI and scripting practices
Target Audience
- AI professionals focused on model deployment strategies
- Machine learning specialists seeking to utilize hardware acceleration
- Deep learning engineers developing inference-based solutions
