BASE-I

Benchmarking for AI for Science at Exascale (BASE – I) was the first phase preceding a more challenging and more extensive project BASE-II. Dr. J. Thiyagalingam (Scientific Computing Department, STFC) was the principal investigator who ran the project, with the help of co-investigators:

Prof W.G. Armour (Engineering Science Department, University of Oxford)
Dr. T. Snow (Science Division, Diamond Light Source)
Dr. T. Lazauskas (Research Department, The Alan Turing Institute)
Dr. J. Zuntz (Sch of Physics and Astronomy, University of Edinburgh)
Dr. T. Burnley (Scientific Computing Department, STFC)
Prof O. Lahav (Physics and Astronomy department, UCL)

Outcomes

BASE-I has resulted in substantial success with the following key outcomes:



Release of the world’s first AI for Science benchmark suite with a framework – SciML-Bench. To the best of our knowledge, this is the first initiative aimed at Benchmarking for the AI for Science community with tangible software delivery.



A series of community events, workshops, and meetings with a number of key international partners, covering a broad spectrum of domains, and user groups, from scientists to RSEs to experts from different parts of the world.



Increased interest in the UK-led benchmarking efforts from various US and international communities, resulting in the BASE-I team being invited to several benchmarking bodies (including the MLCommons Science Working Group, International Bench Council, and Benchmarking for Science activities at US labs).



Publication of several journal articles.



Overwhelming interest in the BASE-I work and resulting benchmark suite from the scientific community. To date, we have more than 40 benchmark cases from various scientific domains and institutions worldwide.



Support from various partners in the US for mirroring our large datasets



The organisation of a series of training programmes targeting RSEs & scientists.

These outcomes contributed to the Blueprint AI for Science at Exascale project being precisely tailored to the most relevant and up-to-date requirements and challenges of scientific, academic, and industry communities in the UK and worldwide.

Our community