BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T143011Z
LOCATION:B302-B305
DTSTART;TZID=America/New_York:20241120T100000
DTEND;TZID=America/New_York:20241120T170000
UID:submissions.supercomputing.org_SC24_sess533_post155@linklings.com
SUMMARY:Creating Code LLMs for HPC: It’s LLMs All the Way Down
DESCRIPTION:Aman Chaturvedi, Daniel Nichols, Siddharth Singh, and Abhinav 
 Bhatele (University of Maryland)\n\nLarge language models (LLMs) are being
  used increasingly by software developers, researchers, and students to as
 sist them in coding tasks. While newer LLMs have been improving their codi
 ng abilities with regards to serial coding tasks, they consistently perfor
 m worse when it comes to parallelism and HPC-related coding tasks. Bridgin
 g this gap and creating HPC-capable code LLMs could drastically improve th
 e quality and quantity of code research software developers can write. The
  current poor performance of LLMs on HPC-related problems can be partially
  attributed to the lack of significant HPC data in their training, which i
 s what we address in this poster. We present HPC Coder v2, a new LLM creat
 ed by fine-tuning a previous code LLM using HPC synthetic data. We demonst
 rate that it is one of the most capable open-source LLMs for generating pa
 rallel code to date.\n\nRegistration Category: Tech Program Reg Pass, Exhi
 bits Reg Pass\n\nSession Chairs: Ayesha Afzal (Friedrich-Alexander-Univers
 ität Erlangen-Nürnberg, Erlangen National High Performance Computing Cente
 r); Sally Ellingson (University of Kentucky); and Alan Sussman (University
  of Maryland)\n\n
END:VEVENT
END:VCALENDAR