Anthill Inside 2017

On theory and concepts in Machine Learning, Deep Learning and Artificial Intelligence. Formerly Deep Learning Conf.

Participate

PyTorch Demystified, Why Did I Switch

Submitted by Sherin Thomas (@hhsecond) on Sunday, 28 May 2017


Preview video

Section: Full talk Technical level: Beginner

View proposal in schedule

Abstract

PyTorch entered into the realm of DL framework with the promise of being “Numpy on GPU”. The obvious failures of static graph implementation for certain use cases is increasing industry wide adoption of PyTorch. Dynamic Computation Graph being the backbone of PyTorch, comes with some perks.
- Super intuitive - Shallow learning curve - Amazing community and discussion forum - Easy debugging

It puts you explicitly in control of your compute, there’s no compiler trying to be clever and “help you” or speed your code up, but in reality mostly leading to a massive amount of debugging headaches. PyTorch should be the go-to framework for the problem with “dynamicity” in the core, especially for RNNs and highly dynamic Reinforcement Learning algorithms.
My talk would be to convince the audience about the advantages of PyTorch. For what problems pytorch would be the best solution. I will have benchmarks with popular DL frameworks that review speed, performance, memory, and easiness. I’ll try to build on top of a comparison foundation with TF and NumPy so that the audience could correlate with their day to day research/development projects. The demo would be to portray the best cases where pytorch really shines over other DL frameworks.

Note: Code I developed for the presentation would be available under MIT licence

Outline

  • Computational Graphs [5 minutes]
  • Static and Dynamic graphs [5 minutes]
  • Introduction to PyTorch [10 minutes]
    • Variables
    • Functional
    • nn Module
    • Optim
    • AutoGrad
  • Examples [10 minutes]
    • Simple Neural Networks
    • Encoder Decoder Network
    • Stack-augmented Parser-Interpreter Neural Network
  • PyTorch and Caffe2 [2 minutes]
  • How to visualize [5 minutes]
  • Benchmarking [1 minute]
  • QnA

Requirements

  • Basic knowledge about how Neural Networks work
  • Knowledge on Numpy operations is recommended
  • It would help (but not mandatory) if the audience could go through this 60 minutes pytorch tutorial

Speaker bio

I am working as an AI developer in CoWrks. Here we work on a broad spectrum of use-cases in NLP and CV. I am currently focusing on NLP, particularly word representation using image, speech and predefined lexical structure. In internet, I go by hhsecond.

Links

Slides

https://docs.google.com/presentation/d/17PNe19jaJNPCeer0SAmfGjS6Pk310iVc_2nm_HDSjzQ/edit?usp=sharing

Preview video

https://www.youtube.com/watch?v=vduog9FbUFk

Comments

  • Zainab Bawa (@zainabbawa) 2 years ago

    Please submit slides and preview video to complete review of this proposal.

  • Sandhya Ramesh (@sandhyaramesh) 2 years ago

    Hi Sherin, please also submit a two minute preview video. Thanks!

  • Jewel Lukose (@lukojewel) a year ago

    Awesome. All the best Sherin

Login with Twitter or Google to leave a comment