“Computing Access Training” are training paths designed to provide the necessary skills to access IT4LIA services, including access to computing resources, data preparation, processing, model training and inference services.
This type of training is not provided individually but together with other IT4LIA services; it is an integral part of the service package that IT4LIA offers to businesses, start-ups and public administrations, which can be accessed after completing an assessment.
The training pathway spans three levels: the Introductory module covers basics of HPC and AI integration, the Intermediate focuses on workflow optimisation and scaling, while the Advanced explores distributed training, orchestration and MLOps.
Computing access training
To enable access to IT4LIA’s computing resources
Introductory Training Module on “AI on HPC”
Description:
This module offers a structured introduction to integrating AI techniques on HPC infrastructures. It is designed to provide participants with both theoretical and practical insights into the main paradigms, tools, and architectures used in the field of AI on HPC. The module can be customized and may cover the following topics:
- Introduction to the IT4LIA HPC infrastructure
- Job submission to the scheduler
- Project management on HPC
- Support tools (containers, workload managers, libraries)
Target audience:
Start-ups, SMEs, research institutions, universities, public administration
Intermediate Training Module on “AI on HPC”
Description:
This intermediate module is aimed at users already familiar with HPC infrastructures who wish to deepen their technical understanding of efficiently running AI workflows. The focus is on resource usage optimization and large-scale job management.
Possible topics include:
- Advanced job management with SLURM (array jobs, dependencies, priorities)
- Profiling techniques and performance optimization
- Use of advanced containers and virtualized environments
- Resource usage monitoring and debugging on HPC
- Distributed training with multi-GPU and multi-node setups
Target audience:
Start-ups, SMEs, research institutions, universities, public administration
Advanced Training Module on “AI on HPC”
Description:
This advanced module, with flexible duration, targets expert users who want to fully leverage HPC infrastructures for large-scale AI or computational projects. The course delves into parallelization techniques, orchestration of distributed experiments, use of multi-GPU resources, and advanced environment tuning.
Possible topics include:
- Large-scale distributed training (e.g., with PyTorch DDP, DeepSpeed, Lightning, Axolotl)
- Optimization of node-to-node communication (e.g., NCCL, MPI)
- Scalability and benchmarking of AI applications
- Integration of orchestration and MLOps tools
Target audience:
Start-ups, SMEs, research institutions, universities, public administration
Access IT4LIA's Services
We are here to help you leverage HPC for your AI projects with services and training. Contact us for expert guidance and collaboration opportunities.