v2.1

Singularity container for TF2.1

You can use a Singularity container (for installation Debian/Ubuntu systems run sudo apt install singularity-container command). Note that version 3.0 is needed at least!

As a first step download the container form this release. Then, you can run the container on your machine. We recommend to use -H command to override the Python packages you may have installed in your $HOME folder and --bind parameter to bind the folder you are working in.

singularity shell -H $(pwd) --nv --bind . ../../tf-approximate-gpu.sif

This command switches you to the container (the command line should start with Singularity> prefix). You can run two short tests

python -c "import tensorflow as tf; print(tf.__version__)"
# Response: 2.1.0
python /opt/tf-approximate-gpu/test/test_table_approx_conv_2d.py
# Response: Linf Error: 3.072154584060627e-07 (or different negligible number)

Your container is prepared, and you can run your own script inside it.