This is the third and final module in the Aurora Learning Paths series - Accelerate Python Loops with the Intel AI Analytics Toolkit. Ignite performance of common Python constructs by using NumPy powered by oneAPI. This workshop series will go into detail about how to apply key Intel architectural innovations and libraries via the smart application of NumPy aggregations and ufuncs to achieve amazing performance gains. We’ll delve into NumPy powered by oneAPI, NumPy ufuncs and show implementations of NumPy reductions and Data-parallel Extension for Numba* (numba-dpex)! The planned use cases represent variations on the theme of pairwise distance computations by comparing several alternative approaches.
- Apply NumPy constructs as a Python loop replacement strategy that improves readability, maintainability, performs fast on current hardware and readies code for future HW & SW accelerations via NumPy that Intel builds into silicon.
- Apply reductions using numba-dpex to Intel GPUs to accelerate your Python code.
- Understand the value of NumPy to accelerate Python loops