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 Custom Operator Development
- Why build custom operators? Use cases and constraints.
- CANN runtime structure and operator integration points.
- Overview of TBE, TIK, and TVM in the Huawei AI ecosystem.
Using TIK for Low-Level Operator Programming
- Understanding the TIK programming model and supported APIs.
- Memory management and tiling strategy in TIK.
- Creating, compiling, and registering a custom op with CANN.
Testing and Validating Custom Ops
- Unit testing and integration testing of ops in the graph.
- Debugging kernel-level performance issues.
- Visualizing op execution and buffer behavior.
TVM-Based Scheduling and Optimization
- Overview of TVM as a compiler for tensor ops.
- Writing a schedule for a custom op in TVM.
- TVM tuning, benchmarking, and code generation for Ascend.
Integration with Frameworks and Models
- Registering custom ops for MindSpore and ONNX.
- Verifying model integrity and fallback behaviour.
- Supporting multi-operator graphs with mixed precision.
Case Studies and Specialized Optimizations
- Case study: high-efficiency convolution for small input shapes.
- Case study: memory-aware attention operator optimization.
- Best practices in custom op deployment across devices.
Summary and Next Steps
Requirements
- Robust knowledge of AI model internals and operator-level computation.
- Experience with Python and Linux development environments.
- Familiarity with neural network compilers or graph-level optimizers.
Audience
- Compiler engineers working on AI toolchains.
- Systems developers focused on low-level AI optimization.
- Developers building custom ops or targeting novel AI workloads.
14 Hours
