Optimize your AI projects with HPC and multi-GPUs

Do you train artificial intelligence models and want to improve their performance without spending hours manually testing different configurations?

This training course will show you how to optimise the hyperparameters of your AI models and harness the power of GPUs and a supercomputer’s multi-GPU environments.

Course outline:

Understanding hyperparameters and the main methods for optimising them

• Discovering the concept of pruning
• Optimising hyperparameters on a GPU
• Scaling up in a multi-GPU environment
• Analysing model performance
• Comparing and discussing the gains achieved following optimisation
• Putting the concepts into practice through exercises on the ROMEO2025 supercomputer’s GPUs

Why take part?

• Learn how to optimise the hyperparameters of your artificial intelligence models
• Understand the metrics used to evaluate a model’s performance
• Reduce the time needed to find the best configurations
• Make effective use of multi-GPU resources
• Experiment directly on the ROMEO2025 supercomputer

Target audience:

This training course is aimed at engineers and researchers wishing to optimise their models and hyperparameters by harnessing the multi-GPU computing power available on the nodes of a supercomputer such as ROMEO.

Prerequisites:

• Knowledge of Python. If you do not have this, you can take free courses to learn the basics of Python. These courses include:
• Learn the basics of Python – OpenClassrooms
• Get started with Python for data analysis – OpenClassrooms
• Basic knowledge of high-performance computing (HPC). The two courses ‘Introduction to Supercomputers’ and ‘My First Steps on a Supercomputer’ offered by ROMEO cover this prerequisite.
• Understanding the fundamental concepts of deep learning. The ‘Fundamentals of Deep Learning’ course offered by ROMEO covers this prerequisite.

The ‘Accelerate your AI projects with HPC and multi-GPUs’ course offered by ROMEO is recommended but not compulsory.