Tensorflow is a software library, primarily used for machine learning. Here, artificial neural networks are mapped to multidimentional arrays, so-called tensors and calculations on the tensors are represented as a graph. The library is accessible via Python. It is mostly developed by Google, however it is available as free and open-source software under the Apache license. Further information and countless tutorials are available at https://www.tensorflow.org/.
Using Tensorflow
We provide various Tensorflow variants under the GPU software stack, which you get be typing module load GpuModules
. This will load Python 3.7 as well as Tensorflow2 and Keras with GPU support. You may list all available modules in the software stack with module avail
and switch to other alternatives with module switch <module to replace> <module that replaces>
.
All these libraries support GPUs without further action required, and also run on regular HPC nodes without GPU support. Please refer to information on how to write job scripts and additional references for GPU usage for directions on job scripts.