GPU are very suitable for deep learning (DL) purposes, providing significant (sometimes almost two orders of magnitude) speedups compared to CPU. Currently, most of deep learning frameworks support GPU both for training and inference. However, those frameworks are pretty heavy and have many dependencies. I propose to add GPU support to a simple DL framework with a small amount of dependencies.

Organization

Student

Evgenii Zheltonozhskii

Mentors

  • Stefano Fabri
  • Edgar Riba
close

2017