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 CANN and Ascend AI Processors
- Defining CANN and its pivotal role within Huawei’s AI compute stack.
- An overview of Ascend processor architectures, including models 310, 910, and others.
- A summary of supported AI frameworks and the associated toolchain.
Model Conversion and Compilation
- Utilizing the ATC tool for model conversion from TensorFlow, PyTorch, and ONNX.
- Generating and verifying OM model files.
- Addressing unsupported operators and resolving common conversion challenges.
Deployment via MindSpore and Alternative Frameworks
- Deploying models using MindSpore Lite.
- Integrating OM models with Python APIs or C++ SDKs.
- Utilizing the Ascend Model Manager for streamlined operations.
Performance Optimization and Profiling
- Exploring AI Core, memory, and tiling optimization techniques.
- Profiling model execution using dedicated CANN tools.
- Applying best practices to enhance inference speed and resource efficiency.
Error Handling and Debugging
- Identifying common deployment errors and implementing solutions.
- Interpreting logs and utilizing the error diagnosis tool effectively.
- Conducting unit testing and functional validation of deployed models.
Edge and Cloud Deployment Scenarios
- Deploying to Ascend 310 for specialized edge applications.
- Integrating solutions with cloud-based APIs and microservices.
- Examining real-world case studies in computer vision and NLP.
Summary and Future Directions
Requirements
- Proficiency with Python-based deep learning frameworks, including TensorFlow or PyTorch.
- A solid understanding of neural network architectures and model training workflows.
- Foundational knowledge of Linux command-line interfaces (CLI) and scripting.
Target Audience
- AI engineers focused on model deployment strategies.
- Machine learning practitioners aiming to leverage hardware acceleration.
- Deep learning developers constructing efficient inference solutions.
14 Hours
