Get in Touch

Course Outline

Introduction to Huawei CloudMatrix

  • Overview of the CloudMatrix ecosystem and deployment workflow
  • Supported models, formats, and various deployment modes
  • Common use cases and compatible chipsets

Preparing Models for Deployment

  • Exporting models from training tools such as MindSpore, TensorFlow, and PyTorch
  • Utilizing ATC (Ascend Tensor Compiler) for format conversion
  • Distinguishing between static and dynamic shape models

Deploying to CloudMatrix

  • Creating services and registering models
  • Deploying inference services through the UI or CLI
  • Managing routing, authentication, and access controls

Serving Inference Requests

  • Comparing batch and real-time inference flows
  • Implementing data preprocessing and postprocessing pipelines
  • Integrating CloudMatrix services into external applications

Monitoring and Performance Tuning

  • Reviewing deployment logs and tracking requests
  • Managing resource scaling and load balancing
  • Optimizing latency and throughput performance

Integration with Enterprise Tools

  • Connecting CloudMatrix with OBS and ModelArts
  • Leveraging workflows and model versioning
  • Implementing CI/CD for model deployment and rollback strategies

End-to-End Inference Pipeline

  • Deploying a complete image classification pipeline
  • Benchmarking performance and validating accuracy
  • Simulating failover scenarios and system alerts

Summary and Recommended Next Steps

Requirements

  • A solid grasp of AI model training workflows
  • Familiarity with Python-based ML frameworks
  • Fundamental knowledge of cloud deployment concepts

Target Audience

  • AI operations teams
  • Machine learning engineers
  • Cloud deployment specialists utilizing Huawei infrastructure
 21 Hours

Testimonials (2)

Related Categories