First you need a conda environment. Here we are showing how to install miniforge3, a lightweighted conda installation
curl -L -O https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-x86_64.sh
# -b is `batch mode`
bash Miniforge3-Linux-x86_64.sh -b
Then you typically want the conda to automatically setup at login:
~/miniforge3/bin/conda init bash init bash
source ~/.bashrc
Create a new environment and activate it:
ENV_NAME=myenv
conda create --name ${ENV_NAME} -y python=3.9 mamba
conda activate ${ENV_NAME}
if you want make the environment as the default:
cat <<EOF >>~/.bashrc
conda activate ${ENV_NAME}
EOF
Check python is the right one:
which python
# e.g. ~/miniforge3/envs/${ENV_NAME}/bin/python
Avoid conflicts with PIP packages in ~/.local directory (equivalent to PYTHONNOUSERSITE=True
):
conda install -c conda-forge conda-ecosystem-user-package-isolation
Install the environment in the kernel list:
mamba install -y ipykernel
python -m ipykernel install --user --name "${ENV_NAME}" --display-name "${ENV_NAME}"
Check it has been installed:
jupyter kernelspec list
Install the packages that may be needed (this is a pretty comprensive list, customize as needed):
mamba install -y -c conda-forge \
numpy \
h5py \
urllib3 \
pillow \
colorcet \
patchelf \
conda \
boto3 \
pandas \
matplotlib \
scikit-image \
packaging \
jupyter \
notebook \
jupyter_bokeh \
jupyterlab \
jupyter-server-proxy \
jupyter_nbextensions_configurator \
jupyter_contrib_nbextensions \
ipywidgets \
ipympl \
pyviz_comms \
bokeh \
vtk \
voila \
panel
(OPTIONAL) if you have an old version of JupyterLab you may need to run this:
jupyter labextension install @pyviz/jupyterlab_pyviz
Install openvisus:
- NOTE: if you have problems you can install
openvisusnogui
which does not include the OpenVisus viewer
mamba install -c visus -y openvisus
Run the OpenVisus configuration step:
- configuration step will install PyQt5
- will relink OpenVisus binaries to use the PyQt5 version (using
patchelf
internally) - if you are using the
openvisusnogui
version, configure is generally not needed
# first time the configure will cause segmentation fault
python -m OpenVisus configure
python -m OpenVisus configure
# configure is doing:
# python -m pip install --upgrade pip
# python -m pip install numpy
# python -m pip install PyQt5~=5.12.0 PyQtWebEngine~=5.12.0 PyQt5-sip
# ... relinking of openvisus binaries...
Check it is working:
python -c "from OpenVisus import *"
if you want to run a jupyter notebook:
- NOTE: check the status of the firewall (e.g.,
sudo ufw status
) - you may check connectivity using curl (e.g.,
curl -v -L http://<hostname>:8888/?token=XXXX
)
jupyter notebook --ip="$(curl ifconfig.me)" --port 8888 --no-browser --debug \