Learn Data-Parallel Essentials for Python*

Watch this information-packed, two-hour workshop to learn techniques for accelerating AI applications that target Intel® XPUs by using the Python* DPPy library of algorithms and Numba-dpex (Numba Data-Parallel Extension).

Workshop topics include:

• A demonstration of techniques for using the AI Tools to build high-performing Python code that targets Intel CPUs and GPUs.
• An overview of Numba-dpex and data-parallel control (dpctl).
• A demonstration on ways to offload from CPU to GPU when implementing K-means and Pairwise use cases.
• For programmers with SYCL* and GPU programming experience, a code walk-through of how to write an explicit kernel using the @numba_dppy.kernel decorator.
• Several code examples that illustrate data-parallel programming techniques with Python.

Skill level: All

0:00 Introductions

1:30 Numba-dpex agenda

3:30 Introduction to oneAPI

5:28 Introduction to AI Tools

7:30 Intel® VTune™ Profiler

8:27 Intel® Advisor

11:54 Get started with the Intel® Developer Cloud

18:28 Data parallel essentials for Python

21:18 Dpctl

23:28 Compute follows data

25:58 Programming model

26:25 Numba-dpex: Catch up on Q&A

36:04 Automatic offload using the @njit decorator

40:04 Explicit parallel for loop using the @njit decorator

41:10 @dppy.kernel decorator

44:50 Hands-on introduction to Numba-dpex

55:30 Introduction to dpctl

1:12:15 Types of machine learning

1:13:14 Types of supervised learning

1:14:56 Classification and regression

1:15:38 Supervised learning overview

1:16:11 Regression: Numeric answers

1:16:34 Classification: Categorical answers

1:17:39 What is classification?

1:18:12 What is needed for classification?

1:18:51 K-nearest neighbor (KNN) classification example

1:21:42 What is needed to select a KNN model?

1:22:22 Euclidean distance

1:28:20 Manhattan distance

1:28:48 Introduction to patching

1:29:30 Patching alternatives

1:31:38 KNN syntax

1:32:30 Pairwise distance

1:34:07 Cosine distance

1:36:25 Correlation distance

1:40:46 Use @dppy.kernel for pairwise distance

1:42:34 Distance syntax

1:43:40 Hands-on with dpex_Pairwise

1:48:40 Hands-on with dpex_GPAIRS

1:49:48 Example of pairwise distance

1:51:40 Q&A