Opening ceremony

• Alberto Pace (CERN)

Room: 31/3-004 - IT Amphitheatre

A Practical Guide to Modern Natural Language Processing

• Cristian Schuszter (CERN)

Room: 31/3-004 - IT Amphitheatre From its humble beginnings to cutting-edge advancements, this lecture offers a comprehensive yet accessible journey through the history and state-of-the-art techniques in NLP. We’ll explore a few essential concepts, practical applications, and emerging trends without drowning in the complex mathematics needed to unravel these tools. At the end of this lecture, you should be equipped with a general understanding of the potential and inner workings of Natural Language Processing techniques, which you’ll try out in the exercise session.

Functional programming (and why it's relevant for HEP computing)

• Florine de Geus (CERN/University of Twente (NL))

Room: 31/3-004 - IT Amphitheatre Anyone that writes code on a regular basis will likely have come across the term “functional programming”, or perhaps even tried their hand and writing some Haskell functions at some point. Still, its strong theoretical roots and at times unfamiliar syntax can make the initial learning curve quite steep. However, you don’t need to be a mathematical or computer science expert to pick up some essential concepts of the functional programming paradigm. In fact, most likely you are already using some of them in your “normal” imperative programs! In this lecture, we will give a gentle introduction to the world of functional programming with Haskell, and try to convince you how other programming practices, especially those already used in HEP can benefit from "functional thinking".

From Text to Threads: Large Language Models and their impact on the HEP community

• Andrea Valenzuela Ramirez (CERN)

Room: 31/3-004 - IT Amphitheatre Over the past year, Large Language Models (LLMs) have demonstrated remarkable capabilities, with ChatGPT rapidly becoming a buzzword on the Internet. These advanced models have been widely applied in various applications and services, recently appearing in scientific domains as well. In the High Energy Physics (HEP) community, GPT models have appeared at one of the key conferences: the latest Conference on Computing in High Energy and Nuclear Physics. In this lecture, we will examine some foundational components of LLMs, focusing on the well-known GPT models and the techniques to fully leverage their capabilities. We will also review the initial footprints of these technologies within HEP. Starting with the fundamental question: "Can ChatGPT do physics?", we will quickly find that it can already recognize some coding templates from CERN's experiments. Additionally, we will discuss significant drawbacks of these models, such as model hallucinations, which could potentially limit their applicability in rigorous domains. Finally, we will explore the use of LLMs in coding. We will highlight the challenges of using general LLMs for coding while demonstrating prompt engineering strategies designed to turn natural language generation into code generation and understanding.

GPU Programming Made Easy with CuPy

• Bernardo Abreu Figueiredo (Karlsruhe University of Applied Sciences (DE))

Room: 31/3-004 - IT Amphitheatre In scientific environments, Python has become prevalent. At the same time, GPUs have dominated code acceleration use cases in the past years and are used where a large amount of data is processed. This lecture introduces the CuPy library as an easy way to start writing code for GPUs in Python and to accelerate existing applications. The focus is on the capabilities of CuPy through two real-life examples, which demonstrate the versatility of CuPy and its performance improvements for scientific calculations.

Why do machines learn? Introduction to fundamentals and common misconceptions in ML

• Pratik Jawahar (University of Manchester (UK - ATLAS))

Room: 31/3-004 - IT Amphitheatre

Exercise: A Practical Guide to Modern Natural Language Processing

• Cristian Schuszter (CERN)

Room: 513/1-024

The perfectly parallel program: Architectures for hardware acceleration and heterogeneous computing (1/2)

• Zenny Wettersten (CERN)

Room: 31/3-004 - IT Amphitheatre - Technical details of processing units in the context of hardware acceleration - Flynn's taxonomy of computer architectures and its relation to types of software parallelism - Data-level parallelism as a solution to current bottlenecks in scientific computing - Evaluating the possible speedup hardware can provide - How physical limitations in instruction/memory access affect your code

The perfectly parallel program: Design philosophy for parallel programming (2/2)

• Zenny Wettersten (CERN)

Room: 31/3-004 - IT Amphitheatre - Relating the fundamental restrictions and capabilities of hardware parallelism to software engineering - Defining what makes an algorithm "parallel" or not - Considerations for designing a parallel algorithm with respect to control flow and memory access - Legacy code, and how to go about re-designing existing software for novel hardware

Unraveling Grid Computing: From Basics to WLCG

• Robin Hofsaess (KIT - Karlsruhe Institute of Technology (DE))

Room: 31/3-004 - IT Amphitheatre In an era where computational demands surpass the capabilities of individual systems and where collaboration across borders becomes paramount, grid computing is the foundation for today's collaborative research. It enables users to access and analyze experiment data from all around the world without further knowledge of the complex systems in the background. In this lecture, however, we will glance behind the user's single point of entry and delve into the foundational aspects of grid computing, from the historical background to the key concepts of a decentralized modern grid infrastructure. After the key concepts, we will examine the Worldwide LHC Computing Grid (WLCG) - the world's most sophisticated scientific computing grid - as a paradigm of contemporary grid infrastructure.

Exercise: Functional programming (and why it's relevant for HEP computing) (1/2)

• Florine de Geus (CERN/University of Twente (NL))

Room: 513/1-024 Anyone that writes code on a regular basis will likely have come across the term “functional programming”, or perhaps even tried their hand and writing some Haskell functions at some point. Still, its strong theoretical roots and at times unfamiliar syntax can make the initial learning curve quite steep. However, you don’t need to be a mathematical or computer science expert to pick up some essential concepts of the functional programming paradigm. In fact, most likely you are already using some of them in your “normal” imperative programs! In this lecture, we will give a gentle introduction to the world of functional programming with Haskell, and try to convince you how other programming practices, especially those already used in HEP can benefit from "functional thinking".

Exercise: Functional programming (and why it's relevant for HEP computing) (2/2)

• Florine de Geus (CERN/University of Twente (NL))

Room: 513/1-024

Exercise: Unraveling Grid Computing: From Basics to WLCG

• Robin Hofsaess (KIT - Karlsruhe Institute of Technology (DE))

Room: 513/1-024

Computer Networks in HEP (1/2)

• Spyridon Trigazis (CERN)

Room: 31/3-004 - IT Amphitheatre

Computer Networks in HEP (2/2)

• Spyridon Trigazis (CERN)

Room: 31/3-004 - IT Amphitheatre

Generative Machine Learning in HEP: Simulation and beyond

• Francesco Vaselli (Scuola Normale Superiore & INFN Pisa (IT))

Room: 31/3-004 - IT Amphitheatre Generative Machine Learning models are at the forefront of many recent developments in science, with groundbreaking implications. High Energy Physics is no exception, and a wide range of algorithms is already being used to speed-up and improve simulation, monitor data quality and perform anomaly detection. In this lecture, we’ll uncover the hidden mechanisms of these algorithms, show the common building blocks and the key differences, and provide an overview of how this type of machine learning application can pave the way for future physics discoveries. Basic knowledge of machine learning is helpful but not required to follow the lecture. Join us in the linked hands-on session to start applying what you’ve learned to the problem of particle jet simulation, and try to design the best performing model yourself!

Advanced git course: How to git good! (1/2)

• Simone Rossi Tisbeni (Universita e INFN, Bologna (IT))

Room: 31/3-004 - IT Amphitheatre This 2-hour course will teach you how to use Git beyond the basic add, commit, push routine. We'll consolidate core concepts and introduce powerful commands like switch, restore, rebase, and reset. We will also discuss the differences between rebasing vs. merging, and explore advanced admin tool such as filter-repo and hooks. The hands-on exercises will reinforce your learning, focusing on mastering rebasing techniques in a dedicated practice repository. Optional advanced exercises will teach how to write your own git hooks, and use filter-repo to alter the history of the repository.

Advanced git course: How to git good! (2/2)

• Simone Rossi Tisbeni (Universita e INFN, Bologna (IT))

Room: 31/3-004 - IT Amphitheatre This 2-hour course will teach you how to use Git beyond the basic add, commit, push routine. We'll consolidate core concepts and introduce powerful commands like switch, restore, rebase, and reset. We will also discuss the differences between rebasing vs. merging, and explore advanced admin tool such as filter-repo and hooks. The hands-on exercises will reinforce your learning, focusing on mastering rebasing techniques in a dedicated practice repository. Optional advanced exercises will teach how to write your own git hooks, and use filter-repo to alter the history of the repository.

Exercise: Generative Machine Learning in HEP: Simulation and beyond

• Francesco Vaselli (Scuola Normale Superiore & INFN Pisa (IT))

Room: 513/1-024

Intro into Networking for HPC

• Vlad-Andrei Badoiu (University Politehnica of Bucharest)

Room: 31/3-004 - IT Amphitheatre The lecture will provide a brief overview of networking for high-performance computing. The content is divided into four sections: Networking 101. This section delves into the foundational aspects of networking, covering the TCP/IP stack. The Datacenter. This segment offers a succinct walkthrough of a datacenter, explaining the role of a switch, highlighting the two prevalent topologies (dragonfly and fat-trees), and discussing routing and oversubscription. Achieving High Bandwidth and Low Latency. Here, we'll introduce RDMA and RoCE. This part will finish with a discussion on the need for multi path protocols in the datacenter. Current Research. This section will touch upon emerging technologies such as Falcon, EQDS and the targets of the UltraEthernet Consortium.

IPv6. Are we there yet?

• Vlad Nastase (University POLITEHNICA Bucharest)

Room: 31/3-004 - IT Amphitheatre IPv6 has been designed to replace IPv4. Yet, 25 years later, we are still widely using IPv4. This lecture takes a look at the differences between the two protocol versions, how IPv6 is meant to reduce the complexity of IPv4 and some of the reasons adoption of IPv6 has not yet reached 100%.

Closing remarks Room: 31/3-004 - IT Amphitheatre

Exercise: The perfectly parallel program

• Zenny Wettersten (CERN)

Room: 31/3-004 - IT Amphitheatre

Exercise: Computer Networks in HEP

• Spyridon Trigazis (CERN)

Room: 513/1-024

Exercise: Advanced git course: How to git good!

• Simone Rossi Tisbeni (Universita e INFN, Bologna (IT))

Room: 513/1-024