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
Introduction to CANN Optimization Capabilities
- Mechanisms of inference performance management within CANN
- Performance objectives for edge and embedded AI systems
- Insights into AI Core usage and memory distribution
Leveraging the Graph Engine for Analysis
- Fundamentals of the Graph Engine and its execution flow
- Visualization of operator graphs and runtime indicators
- Refining computational graphs to achieve optimization goals
Profiling Tools and Performance Indicators
- Utilizing the CANN Profiling Tool (profiler) for workload evaluation
- Evaluating kernel execution durations and identifying constraints
- Profiling memory access patterns and applying tiling tactics
Developing Custom Operators with TIK
- Overview of TIK and its operator programming framework
- Building custom operators via the TIK DSL
- Validating and benchmarking operator efficiency
Advanced Operator Refinement with TVM
- Basics of TVM integration with CANN
- Automated tuning approaches for computational graphs
- Criteria for selecting between TVM and TIK implementations
Memory Optimization Strategies
- Controlling memory structures and buffer positioning
- Methods for minimizing on-chip memory usage
- Best practices for asynchronous processing and resource reuse
Practical Deployment and Case Studies
- Case analysis: Tuning performance for smart city camera systems
- Case analysis: Optimizing inference stacks for autonomous vehicles
- Protocols for iterative profiling and ongoing enhancement
Requirements
- A solid grasp of deep learning model structures and training processes
- Proficiency in deploying models using CANN, TensorFlow, or PyTorch
- Knowledge of Linux command-line interfaces, shell scripting, and Python programming
Target Audience
- AI performance engineers
- Specialists in inference optimization
- Developers focused on edge AI or real-time system applications