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
- An overview of CANN and its function within Huawei’s AI compute stack.
- A look at Ascend processor architectures, including models 310 and 910.
- An examination of supported AI frameworks and the associated toolchain.
Model Conversion and Compilation
- Utilizing the ATC tool for converting models from TensorFlow, PyTorch, and ONNX.
- The process of creating and verifying OM model files.
- Strategies for addressing unsupported operators and common conversion challenges.
Deploying with MindSpore and Other Frameworks
- Methods for deploying models using MindSpore Lite.
- Integrating OM models via Python APIs or C++ SDKs.
- Utilizing the Ascend Model Manager for efficient workflow management.
Performance Optimization and Profiling
- Insights into AI Core, memory, and tiling optimizations.
- Techniques for profiling model execution using CANN tools.
- Best practices for enhancing inference speed and resource efficiency.
Error Handling and Debugging
- Identification and resolution of common deployment errors.
- Interpreting logs and leveraging error diagnosis tools.
- Conducting unit tests and functional validation for deployed models.
Edge and Cloud Deployment Scenarios
- Deploying applications on Ascend 310 for edge computing.
- Integrating solutions with cloud-based APIs and microservices.
- Case studies in computer vision and NLP applications.
Summary and Next Steps
Requirements
- Proficiency in Python-based deep learning frameworks, including TensorFlow or PyTorch.
- A solid understanding of neural network architectures and standard model training workflows.
- Foundational knowledge of Linux CLI operations and scripting.
Target Audience
- AI engineers focused on model deployment strategies.
- Machine learning practitioners aiming to leverage hardware acceleration.
- Deep learning developers constructing inference solutions.
14 Hours
