Tensorflow

Software Summary

Mesabi

Default Module:

Other Modules Available:

1.3_gpu, 1.3_gpu_python3, 1.4_gpu, 1.4_gpu_python3

Last Updated On:

Tuesday, August 29, 2023

Mesabi K40

Default Module:

Other Modules Available:

Last Updated On:

Mangi

Default Module:

1.4_gpu_python3

Other Modules Available:

1.3_gpu, 1.3_gpu_python3, 1.4_gpu, 1.4_gpu_python3

Last Updated On:

Tuesday, August 29, 2023

Mangi v100

Default Module:

Other Modules Available:

Last Updated On:

NICE

Default Module:

Other Versions Available:

Last Updated On:

Tuesday, August 29, 2023

Support Level:

Secondary Support

Software Access Level:

Open Access

Home Page:

Link

Software Categories:

Numerical Libraries

Statistics

Libraries

Software Description

From the Tensorflow website:

TensorFlow is an open source software library for numerical computation using data flow graphs. Nodes in the graph represent mathematical operations, while the graph edges represent the multidimensional data arrays (tensors) communicated between them. TensorFlow was originally developed by researchers and engineers working on the Google Brain Team within Google's Machine Intelligence research organization for the purposes of conducting machine learning and deep neural networks research, but the system is general enough to be applicable in a wide variety of other domains as well.

Software Documentation

Software Documentation Tabs

General Linux

As a python package, Tensorflow is best installed at MSI using Anaconda. MSI provides certain Tensorflow versions as modules for compatibility and to fill dependencies in other modules, but users wishing to run their own Tensorflow calculations are better served by installing it locally. This offers the advantages that 1) users can select the appropriate version of Tensorflow for their work, and 2) users can also install any other python libraries that are needed.

An example of a worflow for installing Tensorflow locally within a conda environment might look like:

module load python
mamba create -n my_tensorflow python=3 tensorflow -c conda-forge

Then, to use that environment in future sessions, you would run:

module load python
source activate my_tensorflow

You can also use conda to install other packages to your tensorflow environment that you may need. For documentation on managing packages in a conda environment, see the official conda documentation.

If you would like to use Tensorflow on one of our GPU-enabled systems and take advantage of the available CUDA cores, you would need to modify the conda command:

mamba create -n my_tensorflow python=3 tensorflow-gpu -c anaconda

Mesabi

Tensorflow is available for use on MSI's Mesabi k40 nodes. To load tensorflow into your environment, use the following command:

module load tensorflow

Tensorflow is frequently used via its python interface. To use tensorflow in this manner after loading in the module, use the following procedure:

python

>>> import tensorflow as tf

There are a variety of tutorials available for using tensorflow on the official website.