This course is dedicated to the supercomputer LUMI. EuroHPC Joint Undertaking’s LUMI supercomputer reached the third spot on the Top500 list of the world’s fastest supercomputers released at the SC22 conference in Dallas, USA, on 14 November 2022. LUMI reached a measured High-Performance Linpack (HPL) performance of 309 petaflops. This makes LUMI the fastest supercomputer in Europe.
Participants will learn about its architecture and key parameters, how to access LUMI resources, how to run jobs, and how to use the GPU accelerated partition for AI applications.
The course will be delivered in hybrid mode. You can attend the course at IT4Innovations in Ostrava or online.
Mgr. Branislav Jansík, Ph.D. - Supercomputing Services Director of IT4Innovations
Jan Vicherek - LUMI Support Team, IT4Innovations
Georg Zitzlsberger - trainer and researcher at the Advanced Data Analysis and Simulations Lab, IT4Innovations
Matias Jääskeläinen - Training Coordinator, CSC, Finland
9:00 – 9:30 About LUMI (B. Jansík) 30 mins
9:30 – 10:15 Introduction of LUMI supercomputer (J. Vicherek) 45 mins
- Architecture and key parameters
- LUMI-C partition
- LUMI-G partition
- Scratch and Project storage
10:15 – 10:30 Break
10:30 – 11:30 How to access the LUMI supercomputer (B. Jansík + M. Jääskeläinen + J. Vicherek) 30 mins + 30 mins
In this part, the participants will be split into two groups. The access mechanism for users from the Czech Republic will be described in the first and the access mechanisms for users from other EU countries will be discussed in the second group.
- Access mechanisms
- How to submit an application for computational resources
- Evaluation of application
- Account creation request
- First login to the LUMI (guided examples)
- How to run jobs (guided examples)
11:30 – 12:30 Lunch Break
12:30 – 13:30 The use of LUMI supercomputer (J. Vicherek) 60 mins
- Computing environment and available software libraries and tools
- HPC resources allocation, SLURM (guided examples)
- Scratch and Project storages (guided examples)
13:30 – 14:30 Efficient multi-GPU and multi-node execution of Deep Learning frameworks (G. Zitzlsberger) 60 mins
- Introduction to Data Parallel Deep Learning with Horovod
- Multi-node/-GPU aware Data Processing Pipelines
- Demonstration of Multi-node/-GPU Examples using Tensorflow/Keras
14:30 – 15:00 Q&A and closing
This project has received funding from the European High-Performance Computing Joint Undertaking (JU) under grant agreement No 101101903. The JU receives support from the Digital Europe Programme and Germany, Bulgaria, Austria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, Greece, Hungary, Ireland, Italy, Lithuania, Latvia, Poland, Portugal, Romania, Slovenia, Spain, Sweden, France, Netherlands, Belgium, Luxembourg, Slovakia, Norway, Türkiye, Republic of North Macedonia, Iceland, Montenegro, Serbia. This project has received funding from the Ministry of Education, Youth and Sports of the Czech Republic.
This course was supported by the Ministry of Education, Youth and Sports of the Czech Republic through the e-INFRA CZ (ID:90254).