BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20220812T074357Z
LOCATION:Sydney Room
DTSTART;TZID=Europe/Stockholm:20220627T133000
DTEND;TZID=Europe/Stockholm:20220627T153000
UID:submissions.pasc-conference.org_PASC22_sess117@linklings.com
SUMMARY:MS1C - Towards ESM-ML Hybrids with Practical Workflows and Model I
 ntegration
DESCRIPTION:Minisymposium\n\nA promising application area for Machine Lear
 ning (ML) in the domain of Earth System Modelling (ESM) and Numerical Weat
 her Prediction (NWP) revolves around scalable solutions that embed ML in r
 unning ESM/NWP code. For example, in order to ensure the long-term stabili
 ty of ML models that can substitute existing model parameterizations, it m
 ay be required to tune an ML model against a live numerical model, in addi
 tion to the challenge of getting initial training data extracted from the 
 model at possibly high spatial and temporal resolutions in the first place
 . Such setups are much more complex than offline training settings in term
 s of software integration and CPU-GPU data exchange in an HPC setup. The s
 tandard ML workflows will generate models for inference whose utility is c
 urrently limited to the Python ecosystem. Exporting them to work with ESMs
  in Fortran or C environments results in cumbersome workflows. Tackling su
 ch challenges is of key value to research and development efforts that aim
  to build ESM-ML hybrids. To encourage collaboration and formulation of be
 st practices, the minisymposium will bring together ESM and ML developers,
  HPC experts as well as software engineering expertise revolving around bu
 ilding Python/Fortran/C bridges, and foster exchange with other applicatio
 n domains.\n\nInfero: An API for Machine Learning Inference in Operations\
 n\nBonanni, Chantry, Hawkes, Dueben, Quintino...\n\nMachine Learning model
 s are often assembled and trained within a python environment. In most cas
 es, inference is also run in the same environment, and no additional progr
 amming effort is required. Nevertheless, there are cases when the trained 
 machine learning model must be used for inference within...\n\n-----------
 ----------\nOnline Machine Learning for ESM: Using PyTorch and TensorFlow 
 in Fortran with SmartSim\n\nRigazzi\n\nThe inclusion of Machine Learning (
 ML) into Earth System Models (ESM) presents unique domain science and engi
 neering challenges. SmartSim is an open source library dedicated to addres
 sing the engineering challenges around integrating ML into ESM, and other 
 HPC simulations like ESMs, at scale. SmartS...\n\n---------------------\nF
 ortran-Pytorch Bridge for Machine Learning Applications in Earth System Mo
 dels\n\nArnold, Sharma\n\nIn recent years, machine learning (ML) based par
 ameterizations have become increasingly common in Earth System Models (ESM
 ). They can be computationally faster and allow for the inclusion of compu
 tationally intensive parameterizations for sub-grid scale physical process
 es. In our application, we wan...\n\n---------------------\nDiscussion: Sh
 aring Experiences and Aligning Practices\n\nWeigel\n\nThe last slot of the
  symposium will be a discussion on the available frameworks/solutions, cur
 rent use cases and projects in progress on combining different ESMs with M
 L and how to come to best practices, seek alignment between different effo
 rts and share experiences using some of the solutions pres...\n\n\nDomain:
  Climate, Weather and Earth Sciences, Computer Science and Applied Mathema
 tics
END:VEVENT
END:VCALENDAR