Calendar

DESCRIPTION: This course introduces Artificial Intelligence (AI), a science focusing on developing intelligent systems capable of autonomous behavior. In this course, we explore the exciting world of AI, introducing its definition and history. We discuss the advantages and challenges of AI at present, along with various applications and projects that demonstrate its capabilities. Throughout the session, participants will gain insights into different types of AI, learn about running predefined projects, and discover AI showcases on various platforms. By the end of the course, participants will have the knowledge and resources to start their own AI projects with their data and explore the latest AI advancements in our clusters.

TEACHER: Nastaran Shahparian (SHARCNET, York University)

LEVEL: Introductory

FORMAT: Lecture

CERTIFICATE: Attendance and Completion

PREREQUISITE: Basic Python knowledge and know-how is beneficial but not required.

DESCRIPTION: Python is a popular language because it is easy to create programs quickly with simple syntax and a "batteries included" philosophy. However, there are some drawbacks to the language. It is notoriously difficult to parallelize because of a component called the global interpreter lock, and Python programs typically take many times longer to run than compiled languages such as Fortran, C, and C++, making Python less popular for creating performance-critical programs. Dask was developed to address the first problem of parallelism. The second problem of performance can be addressed by either using modules already compiled into fast C/C++ code, such as NumPy, or by converting performance-critical parts into a compiled language such as C/C++ nearly automatically using Cython. Together Cython and Dask can be used to gain greater performance and parallelism of Python programs.

Other than having some prior experience with a programming language, preferably Python, this is a beginner level course. During the course we will program together to build out a script used to demonstrate course concepts. This will take slightly longer than half the time, while hands on exercise will use the remaining time. No Alliance account is required.

TEACHER: Chris Geroux (ACENET, Dalhousie University)

LEVEL: Introductory

FORMAT: Lecture + follow along coding + hands on exercises

CERTIFICATE: Attendance

PREREQUISITES: Should have experience programming in at least one language, ideally Python.

DESCRIPTION: Larger Deep Neural Networks (DNNs) are typically more powerful, but training models across multiple GPUs or multiple nodes isn't trivial and requires a an understanding of both AI and high-performance computing (HPC). In this workshop we will give an overview of activation checkpointing, gradient accumulation, and various forms of data and model parallelism to overcome the challenges associated with large-model memory footprint, and walk through some examples.

TEACHER: Jonathan Dursi (NVIDIA)

LEVEL: Intermediate/Advanced

FORMAT: Lecture + Demo

CERTIFICATE: Attendance

PREREQUISITES:

* Familiarity with training models in Pytorch on a single GPU will be assumed.

DESCRIPTION: We will cover how to write optimized code in C, and how to include this into your Python code. We will look at Cython, as well as pure C. If time permits, we will also look at including FORTRAN.

TEACHER: Joey Bernard (ACENET, University of New Brunswick)

LEVEL: Intermediate

FORMAT: Lecture + Hands-on

CERTIFICATE: Attendance

PREREQUISITES:

* Basic Python programming experience.
* One knows how to use a C compiler.