BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//pyconde-pydata-berlin-2023//talk//GXAKV8
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-pyconde-pydata-berlin-2023-GXAKV8@pretalx.com
DTSTART;TZID=CET:20230417T105000
DTEND;TZID=CET:20230417T113500
DESCRIPTION:When building PyTorch models for custom applications from scrat
 ch there's usually one problem: The model does not learn anything. In a co
 mplex project\, it can be tricky to identify the cause: Is it the data? A 
 bug in the model? Choosing the wrong loss function at 3 am after an 8-hour
  coding session?\n\nIn this talk\, we will build a toolbox to find the cul
 prits in a structured manner. We will focus on simple ways to ensure a tra
 ining loop is correct\, generate synthetic training data to determine whet
 her we have a model bug or problematic real-world data\, and leverage pyte
 st to safely refactor PyTorch models. \n\nAfter this talk\, visitors will 
 be well equipped to take the right steps when a model is not learning\, qu
 ickly identify the underlying reasons\, and prevent bugs in the future.
DTSTAMP:20260608T021128Z
LOCATION:B05-B06
SUMMARY:Honey\, I broke the PyTorch model >.< - Debugging custom PyTorch mo
 dels in a structured manner - Clara Hoffmann
URL:https://pretalx.com/pyconde-pydata-berlin-2023/talk/GXAKV8/
END:VEVENT
END:VCALENDAR