HydroCS Wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== How to set up TensorFlow in a micromamba environment ======

===== Using micromamba =====

<code bash>
mm create -p ./env -y tensorflow
mm activate ./env

# executable stack needed to be cleared on Slackware
patchelf --clear-execstack $CONDA_PREFIX/lib/libtensorflow_cc.so.2

# make sure GPU is recognized
python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
</code>

===== Using PyPI =====

<code bash>
# the latest tensorflow supports python 3.13
mm create -p ./env -y python=3.13
mm activate ./env

# add cuda lib paths to LD_LIBRARY_PATH automatically when activating the
# environment
mkdir -p $CONDA_PREFIX/etc/conda/activate.d
cat > $CONDA_PREFIX/etc/conda/activate.d/tensorflow-cuda.sh <<'EOF'
export _OLD_LD_LIBRARY_PATH="$LD_LIBRARY_PATH"

export LD_LIBRARY_PATH="$(python - <<'PY'
import site
from pathlib import Path

for sp in site.getsitepackages():
    nvidia = Path(sp) / "nvidia"
    if nvidia.exists():
        for lib in nvidia.glob("*/lib"):
            print(lib, end=":")
PY
)$LD_LIBRARY_PATH"
EOF

mkdir -p $CONDA_PREFIX/etc/conda/deactivate.d
cat > $CONDA_PREFIX/etc/conda/deactivate.d/tensorflow-cuda.sh <<'EOF'
export LD_LIBRARY_PATH="$_OLD_LD_LIBRARY_PATH"
EOF

pip install tensorflow[and-cuda]

. $CONDA_PREFIX/etc/conda/activate.d/tensorflow-cuda.sh

# make sure GPU is recognized
python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
</code>