TACC Machine Learning Containers

Containers for running ML applications on TACC GPU sysems

Image information

All images are hosted on Docker Hub

Please refer to our system and version tables which support the following base operating systems:

• Centos7
• Ubuntu 20.04

Usage

$ module load tacc-singularity
$ singularity pull docker://tacc/tacc-ml:ubuntu20.04-cuda11-tf2.6-pt1.10
$ singularity exec --nv tacc-ml_ubuntu20.04-cuda11-tf2.6-pt1.10.sif python -c 'import tensorflow as tf; print(tf.test.is_gpu_available())'                                                           

Development

Begin FROM an image that supports your target TACC system, and add any necessary packages from there.

FROM tacc/tacc-ml:ubuntu20.04-cuda11-tf2.6-pt1.10

RUN conda install new_package

Once you are finished building your new container, push it to dockerhub and then pull to TACC.

Centos7 Images

			System
CUDA	TensorFlow	PyTorch	Maverick2	Frontera/RTX	Longhorn	Lonestar6
11	2.6	1.10	X	X		X
11	2.7	1.10			X

centos7-cuda11-tf2.6-pt1.10

• Dockerfile
• URL: tacc/tacc-ml:centos7-cuda11-tf2.6-pt1.10

ppc64le-centos7-cuda11-tf2.7-pt1.10

• Dockerfile
• URL: tacc/tacc-ml:ppc64le-centos7-cuda11-tf2.7-pt1.10

Ubuntu20.04 Images

			System
CUDA	TensorFlow	PyTorch	Maverick2	Frontera/RTX	Longhorn	Lonestar6
11	2.6	1.10	X	X		X
11	2.7	1.10			X

ubuntu20.04-cuda11-tf2.6-pt1.10

• Dockerfile
• URL: tacc/tacc-ml:ubuntu20.04-cuda11-tf2.6-pt1.10

ppc64le-ubuntu20.04-cuda11-tf2.7-pt1.10

• Dockerfile
• URL: tacc/tacc-ml:ppc64le-ubuntu20.04-cuda11-tf2.7-pt1.10101010101010101010