Skip to content
Type to start searching

    Table of Contents

    TensorFlow

    Deep learning framework for Python.

    News

    TensorFlow is now available on Mahti with GPU support in the new Mahti-AI partition.

    All TensorFlow installations on Mahti will be based on Singularity. This is mainly due to the performance issues of conda-based environments on shared file systems, causing long start-up delays for Python scripts on CSC's supercomputers. For more information, see the Singularity section in our machine learning guide.

    Available

    Currently supported TensorFlow versions:

    Version Module Puhti Mahti Environ. Horovod Notes
    2.5.0 tensorflow/2.5 X X Sing. X
    2.4.1 tensorflow/2.4 X X Sing. X
    2.4.0 tensorflow/2.4-hvd X - Conda X
    2.4.0 tensorflow/2.4-sng X - Sing. -
    2.3.1 tensorflow/nvidia-20.12-tf2-py3 X - Sing. -
    2.3.0 tensorflow/2.3 X - Sing. -
    2.2.0 tensorflow/nvidia-20.07-tf2-py3 X - Conda X experimental Horovod support
    2.2.0 tensorflow/2.2-hvd X - Conda X
    2.2.0 tensorflow/2.2 X - Sing. -
    2.1.0 tensorflow/nvidia-20.03-tf2-py3 X - Sing. -
    2.1.0 tensorflow/nvidia-20.02-tf2-py3 X - Sing. -
    2.0.0 tensorflow/nvidia-19.11-tf2-py3 X - Sing. -
    2.0.0: tensorflow/2.0.0 X - Conda -
    2.0.0 tensorflow/2.0.0-hvd X - Conda X
    1.15.0 tensorflow/1.15-hvd X - Conda X
    1.14.0: tensorflow/1.14.0 X - Conda -
    1.14.0 tensorflow/1.14.0-cpu X - Conda - Optimized for CPU
    1.13.1: tensorflow/1.13.1 X - Conda -
    1.13.1 tensorflow/1.13.1-hvd X - Conda X

    Includes TensorFlow and Keras with GPU support via CUDA.

    Modules starting with nvidia are based on NVIDIA's optimized container images from NGC with some CSC specific additions. See NVIDIA's TensorFlow container release notes for more information on provided software versions.

    If you find that some package is missing, you can often install it yourself with pip install --user. If you think that some important TensorFlow-related package should be included in a module provided by CSC, you can send an email to servicedesk@csc.fi.

    Some modules are Singularity-based (indicated in the "Environ." column in the table above). Wrapper scripts have been provided so that common commands such as python, python3, pip and pip3 should work as normal. For more information, see the Singularity section in our machine learning guide.

    Some modules support Horovod, which is our recommended framework for multi-node jobs, i.e., jobs needing more than 4 GPUs. For more information, read the Multi-GPU section in our machine learning guide.

    License

    TensorFlow is licensed under Apache License 2.0.

    Usage

    To use this software on Puhti or Mahti, initialize it with:

    module load tensorflow

    to access the default version, or if you wish to have a specific version (see above for available versions):

    module load tensorflow/2.4

    Please note that the module already includes CUDA and cuDNN libraries, so there is no need to load cuda and cudnn modules separately!

    This command will also show all available versions:

    module avail tensorflow

    To check the exact packages and versions included in the loaded module you can run:

    list-packages

    Note

    Note that Puhti login nodes are not intended for heavy computing, please use slurm batch jobs instead. See our instructions on how to use the batch job system.

    Example batch script

    Example batch script for reserving one GPU and 1/4 of the available CPU cores in a single node:

    Puhti

    #!/bin/bash
#SBATCH --account=<project>
#SBATCH --partition=gpu
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=10
#SBATCH --mem=64G
#SBATCH --time=1:00:00
#SBATCH --gres=gpu:v100:1

module load tensorflow/2.4
srun python3 myprog.py <options>

    Mahti

    #!/bin/bash
#SBATCH --account=<project>
#SBATCH --partition=gpusmall
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=32
#SBATCH --time=1:00:00
#SBATCH --gres=gpu:a100:1

module load tensorflow/2.4
srun python3 myprog.py <options>

    Note

    Please do not read a huge number of files from the shared file system, use fast local disk or package your data into larger files instead! See the Data storage section in our machine learning guide for more details.

    Big datasets, multi-GPU and multi-node jobs

    Please see our tutorial for GPU-accelerated machine learning, which covers more advanced topics, including how to work with big data sets, multi-GPU and multi-node jobs.

    More information

    Last edited Thu Jun 3 2021