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 21 hours
Course Outline
Introduction to the Chinese AI GPU Landscape
- Comparison of Huawei Ascend, Biren, and Cambricon MLU
- Contrasting CUDA with CANN, Biren SDK, and BANGPy paradigms
- Market trends and vendor ecosystems
Migration Preparation
- Evaluating the existing CUDA codebase
- Selecting target platforms and SDK revisions
- Setting up toolchains and development environments
Code Adaptation Methods
- Translating CUDA memory operations and kernel logic
- Mapping computational grid and thread structures
- Options for automated versus manual translation
Platform-Specific Solutions
- Leveraging Huawei CANN operators and custom kernels
- The Biren SDK conversion workflow
- Reconstructing models using BANGPy (Cambricon)
Cross-Platform Validation and Refinement
- Monitoring execution across each target platform
- Memory adjustment and parallel processing analysis
- Performance monitoring and iterative improvement
Oversight of Mixed GPU Setups
- Hybrid deployments involving multiple architectures
- Contingency strategies and hardware detection
- Abstraction layers to ensure code longevity
Practical Examples and Standards
- Transferring vision/NLP models to Ascend or Cambricon
- Upgrading inference pipelines on Biren clusters
- Resolving version conflicts and API discrepancies
Conclusion and Future Actions
Requirements
- Background in developing with CUDA or GPU-centric applications
- Grasp of GPU memory structures and computational kernels
- Knowledge of AI model rollout or acceleration processes
Target Learners
- GPU developers
- System architects
- Migration experts
