X

Oracle Linux, virtualization , Enterprise and Cloud Management Cloud technology musings

  • February 7, 2018

Using a BareMetal GPU shape in Oracle Cloud Infrastructure with Oracle Linux 7 and TensorFlow

A lot of developers are using TensorFlow for Machine Learning these days. In Oracle Cloud Infrastructure we provide some great GPU options. One of them is the BM.GPU2.2 shape which is an X7-based GPU system (contains 2 P100 Nvidia GPUs).

When you create an OCI instance using this shape with Oracle Linux 7, it comes pre-installed with the kernel modules to enable the GPUs. Ready to use.

Getting TensorFlow installed is very easy:

Install some prerequisite RPMs, some come from the EPEL yum repo which we provide as part of Oracle Linux and is enabled by default in your yum.repos file.

# sudo yum -y install python-pip python-devel atlas atlas-devel gcc-gfortran openssl-devel libffi-devel

# sudo pip install --upgrade virtualenv

# virtualenv --system-site-packages ~/venvs/tensorflow

# source ~/venvs/tensorflow/bin/activate

Now you can install TensorFlow using pip. use tensorflow-gpu if you want the GPU enabled version otherwise just use tensorflow.

(tensorflow) # pip install --upgrade tensorflow-gpu

or

(tensorflow) # pip install --upgrade tensorflow

To use tensorflow-gpu you have to install the Nvidia CUDA packages. This version of tensorflow depends on version 9.0

(tensorflow) # sudo yum -y install cuda-9-0

Run a TF example:

(tensorflow) #  pip install pandas

(tensorflow) # sudo yum -y install git

(tensorflow) # mkdir git

(tensorflow) # cd git

(tensorflow) # git clone https://github.com/tensorflow/models

(tensorflow) # cd models/samples/core/get_started/

(tensorflow) # python premade_estimator.py

and that's it. Super easy without any manual downloads.

this is a test

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.