University Homepage School Homepage School Contacts School Search

MLP 2017-18  |  News Archive  |  Lectures  |  Labs  |  Group Project  |  Coursework  |  Feedback  |  Computing  |  Piazza  |  Github

Machine Learning Practical (MLP) 2017-18: Computing

The MLP GPU cluster

The MLP cluster consists of a number of rack-mounted servers, each equipped with 8 NVIDIA 1060 GTX 6GB GPUs. From 1 February 2018, there are 8 servers (64 GPUs) available for use (this is one rack), during the first half of February two further racks will become available - a total of 25 servers (200 GPUs). The system is managed using the open source cluster management software Slurm. Slurm has various advantages over the competition, including full support of GPU resource scheduling.

Getting started with the cluster

• Notes and sample code for the cluster and TensorFlow in the 2017-8/semester_2_materials branch of the MLP github.
  • Overview of code in the repo; getting set up within Conda - README.md
  • Brief guide to getting started on the GPU system - notes/gpu-cluster-quick-start.md
  • Basic script to use with Slurm and GPUs (discussed in the above note) - gpu_cluster_tutorial_training_script.sh
  • Some code for using Cifar and the million song data sets (prepared last year), useful if you are using those datasets - cifar10_100_and_million_song_datasets.md
  • Some example python code using TensorFlow

Git

• Github Git Handbook
• Atlassian git tutorials -- Setting up a repisitory, Collaborating
• Atlassian git cheatsheet

---

Copyright (c) University of Edinburgh 2015-2018
The MLP course material is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License
licence.txt
The software at https://github.com/CSTR-Edinburgh/mlpractical is licensed under the Modified BSD License.
This page maintained by Steve Renals.
Last updated: 2018/08/14 16:34:18UTC