Systems for ML

Workshop on Systems for ML at NeurIPS 2019

December 13, 2019

Schedule

8:30-8:40:     Opening Remarks, Aparna Lakshmiratan, Facebook AI


Session 1, Chair: Aparna Lakshmiratan, Facebook AI

Link to recording

8:40 - 9:10     Keynote 1: Machine Learning Reproducibility: An update from the NeurIPS 2019 Reproducibility Co-Chairs, Joelle Pineau, McGill University and Facebook

9:10 - 9:30     Contributed Talk: SLIDE : Training Deep Neural Networks with Large Outputs on a CPU faster than a V100-GPU

9:30 - 9:50     Contributed Talk: NeMo: A Toolkit for Building AI Applications Using Neural Modules


Morning Poster Session, Chair: Sid Sen, Microsoft Research

9:50 - 10:00:     Poster Overview

10:00 - 11:10     Break and Poster Session


Session 2: Chair: Sid Sen, Microsoft Research

Link to recording

11:10 - 11:40     Keynote 2: Balancing Efficiency and Flexibility for DNN Acceleration, Vivienne Sze, MIT

11:40 - 12:00     Contributed Talk: 5 Parallel Prism: A Topology for Pipelined Implementations of Convolutional Neural Networks Using Computational Memory


12:00 - 1:30     Lunch


Systems Bonanza, Chair: Dan Crankshaw, Microsoft Research

Link to recording

1:30 - 3:30     Systems Bonanza (10 minutes each)

  • PyTorch (Dmytro Dzhulgakov, Facebook)
  • TensorFlow (Jonathan Hseu, Google Brain)
  • Keras (Yifei Feng, Google Brain)
  • TVM (Tianqi Chen, OctoML)
  • Ray (Robert Nishihara, UC Berkeley)
  • ONNX Runtime (Faith Xu, Microsoft)
  • CoreML (Reza Farhadi, Apple)
  • Flux (Shashi Gowda, MIT)
  • MLFlow (Matei Zaharia, Stanford University and Databricks)
  • MLPerf (Peter Mattson, Google Brain)
  • Microsoft RL Systems (Paul Mineiro, Microsoft Research)
  • MXNet (Alex Smola, Amazon Web Services)

Afternoon Poster Session

3:30 - 4:30     Break and Poster Session


Session 3, Chair: Sarah Bird, Microsoft Research

Link to recording

4:30 - 5:00     Keynote 3: Programming the Graphcore IPU, Ryota Tomioka, Microsoft Research

5:00 - 5:20     Contributed Talk: LISA: Towards Learned DNA Sequence Search


5:20 - 5:30     Closing Remarks