During the session, I will explain what should be done with models after you’ve finished their training. Such steps will be considered:

  1. Various networks speed optimization approaches
    • pruning
    • quantizing
    • distillation
    • XNORing
  2. Infrastructure preparation
    • graph “freezing”
    • handsome frameworks for deployment
    • what exactly and how should be measured
  3. Another tips and tricks

Initially, the session was built as talk, but also it can be refactored as a workshop.


Illarion Khlestov

Computer Vision specialist mainly focused on Deep Learning approaches. At Ring Ukraine, Illarion has developed and improved a variety of neural networks for computer vision such as classification, detection and video description. He has also worked on time signal processing systems in the past. Illarion has hands-on experience with developing complex machine learning pipelines and moving them to production grade. He believes that someday computers might reach human-level intelligence and we should help them with this.

Event Timeslots (1)

Workshop Room A
-
Illarion Khlestov