BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//cfp.devcon.org//devcon8//talk//ZWHMKX
BEGIN:VTIMEZONE
TZID:IST
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1
TZNAME:IST
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-devcon8-ZWHMKX@cfp.devcon.org
DTSTART;TZID=IST:20261105T140000
DTEND;TZID=IST:20261105T141500
DESCRIPTION:Training an AI model can take minutes\, hours or days. Waiting 
 until the end to inspect a log file is often too late: loss may already be
  diverging\, gradients may be behaving unexpectedly\, validation performan
 ce may have stopped improving\, or a data pipeline may simply be producing
  the wrong results.\n\nVisdom provides an open-source way to see what is h
 appening while an experiment is running.\n\nIn this session\, Norbert Prei
 ning will show how Python applications and machine-learning workflows can 
 send live data to Visdom and turn a running experiment into an interactive
  dashboard. Starting with only a few lines of Python\, we will look at str
 eaming loss and accuracy curves\, NumPy data\, images and model outputs di
 rectly into the browser.\n\nThe session will then move from individual plo
 ts to complete ML workflows: monitoring PyTorch and PyTorch Lightning trai
 ning\, collecting TensorFlow/Keras metrics\, organizing experiments\, reco
 rding hyperparameters and comparing runs. We will also look at how Visdom 
 can be used to inspect images\, embeddings and other intermediate results 
 while a model is still training.\n\nA key idea behind Visdom is that visua
 lization should remain separate from the computation itself. Your training
  framework computes the metrics\; Visdom receives\, stores and visualizes 
 them. This keeps the integration small\, makes Visdom usable with many dif
 ferent frameworks and allows developers to run the visualization server lo
 cally\, on their own infrastructure or alongside remote training systems.\
 n\nThe latest Visdom work also makes this approach useful for larger exper
 iment workflows: experiments can carry parameters\, metrics and tags\, run
 s can be searched and compared\, and live results can be explored through 
 hyperparameter and experiment views.\n\nThe talk will use short code examp
 les and live visualizations to show how quickly a normal Python or ML trai
 ning loop can become observable.\n\nWhether you are training a PyTorch mod
 el\, working with TensorFlow/Keras\, exploring NumPy data or building your
  own AI pipeline\, the goal is the same:\n\n**don't wait for the experimen
 t to finish to understand what it is doing.**
DTSTAMP:20261101T041833Z
LOCATION:Community Led Stage
SUMMARY:Watch Your Models Learn: Real-Time AI/ML Experiments with Visdom - 
 Norbert Preining
URL:https://cfp.devcon.org/devcon8/talk/ZWHMKX/
END:VEVENT
END:VCALENDAR