Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
Ansible Playbook to set up Tensorflow and Keras on Ubuntu 16.04
---
- name: Install Tensorflow and Keras on ubuntu1604
hosts: all
become: yes
handlers:
- include: roles/common/handlers/main.yml
vars:
- cuda_network_installer: http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_9.0.176-1_amd64.deb
- cuda_apt_key: http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub
- cudnn_runtime: https://<MY OWN MIRROR HOST>/libcudnn7_7.0.5.15-1%2Bcuda9.0_amd64.deb
tasks:
- name: Install required packages
apt: pkg={{ item }} state=installed
with_items:
- gcc
- g++
- gfortran
- build-essential
- linux-image-generic
- python3-dev
- python3-pip
- python3-virtualenv
- python3-venv
- libblas-dev
- liblapack-dev
- libatlas-base-dev
- software-properties-common
- pkg-config
- apt-utils
- liblapack-dev
- libfreetype6
- libfreetype6-dev
- libpng12-dev
- libyaml-dev
- g++
- libffi-dev
- libzmq-dev
- libzmq1
- graphviz
- python3-numpy
- swig
- libcurl3-dev
- name: Install kernel headers
apt:
name=linux-headers-{{ ansible_kernel }}
state=present
update_cache=yes
- name: Install CUDA network installer
apt:
deb={{ cuda_network_installer }}
- name: Add Nvidia's repo key.
apt_key:
url={{ cuda_apt_key }}
state=present
- name: Install CUDA packages (1.2G+)
apt:
pkg={{ item }}
update_cache=yes
state=installed
with_items:
- cuda-9-0
- name: Install cuDNN
apt:
deb={{ item }}
with_items:
- "{{ cudnn_runtime }}"
- name: Set up work environment
hosts: all
tasks:
- name: Setup Python environment
pip: name={{ item }}
virtualenv='{{ ansible_env.HOME }}/env'
virtualenv_command='/usr/bin/python3 -m venv'
with_items:
- jupyterlab
- 'tensorflow==1.5'
- 'tensorflow-gpu==1.5'
- keras
- h5py
- pydot
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment