BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Training for Deployment: Methods for Small and Efficient NLP - Ale
 xander Rush\, Cornell Tech
DTSTART:20210520T140000Z
DTEND:20210520T150000Z
UID:TALK160600@talks.cam.ac.uk
CONTACT:Marinela Parovic
DESCRIPTION:Natural language models for translation and classification wor
 k relatively well\, or at least well enough that there is demand for wides
 pread use in real systems. Models developed for research however do not na
 turally translate to deployment scenarios\, particularly on resource const
 rained devices like mobile phones. In this talk I will discuss two axes th
 at make it difficult to deploy NLP models in practice: a) Serial generatio
 n in translation models makes them difficult to optimize\, and b) Fine-tun
 ed parameter size in classification makes models difficult to deploy to en
 d-users. I propose two approaches that aim to circumvent these issues\, an
 d discuss some practical work on deploying large NLP models on edge device
 s. 
LOCATION:https://cam-ac-uk.zoom.us/j/97599459216?pwd=QTRsOWZCOXRTREVnbTJBd
 XVpOXFvdz09
END:VEVENT
END:VCALENDAR