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
Overview of CANN Architecture and Ascend AI Processors
- Definition of CANN and its function within Huawei’s artificial intelligence computing infrastructure
- Description of the Ascend processor architecture, including series such as the 310 and 910
- Summary of compatible AI frameworks and associated development toolchains for government use cases
Model Conversion and Compilation Procedures
- Utilization of the Ascend Toolkit Compiler (ATC) to convert models from TensorFlow, PyTorch, or ONNX formats
- Generation and verification of OM model files
- Management of unsupported operators and resolution of common conversion challenges
Implementation with MindSpore and Compatible Frameworks
- Deployment of models using MindSpore Lite
- Integration of OM models via Python application programming interfaces or C++ software development kits
- Operations involving the Ascend Model Manager
Performance Optimization and System Profiling
- Analysis of AI Core performance, memory management, and tiling optimizations
- Execution profiling of model operations using CANN diagnostic tools
- Recommended practices for enhancing inference latency and resource efficiency in government environments
Error Management and Troubleshooting
- Identification of frequent deployment errors and corresponding corrective actions
- Interpretation of system logs and application of automated error diagnosis utilities
- Execution of unit testing and functional validation for deployed model integrity
Deployment Architectures for Edge and Cloud Environments
- Implementation on Ascend 310 processors for edge computing applications
- Integration with cloud-based application programming interfaces and microservice architectures
- Examination of operational case studies in computer vision and natural language processing domains
Conclusion and Subsequent Action Items
Requirements
- Practical experience utilizing Python-based deep learning environments, including TensorFlow or PyTorch, for government applications.
- Comprehensive knowledge of neural network architectures and standard procedures for model training within federal workflows.
- Foundational proficiency in Linux command-line interfaces and automation scripting.
Target Audience
- AI engineering professionals responsible for deploying model solutions.
- Machine learning specialists focused on optimizing hardware acceleration for government systems.
- Deep learning developers constructing inference architectures.
14 Hours
