Skip to content

Latest commit

 

History

History
114 lines (77 loc) · 3.29 KB

manual-installation.md

File metadata and controls

114 lines (77 loc) · 3.29 KB

Install environment

Tested environment

List of environments that have been tested, and validated where possible.

Get repository

git clone https://gitlab.imt-atlantique.fr/edito/4dvarnet-turbiditymapping.git
cd 4dvarnet-turbiditymapping

Conda configuration

Check if conda is available

Display conda version

conda --version

If a version is displayed, you can directly go to "Install dependencies".

If not, you need to install conda in your local laptop or homedir in a server, to manage python environment.

### Install conda

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
chmod +x Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh -p $HOME/miniconda3 
# Done with installing conda here!

After this a directory will be created to contain all library of conda

### Check Conda environment

Now let’s check whether we have conda or not:

bash (this is to change to Bash mode)
conda --version in case we want to check the version
eval "$(/home/local-user/miniconda3/condabin/conda shell.bash hook)" (please change the directory to yours)
conda init

Install conda on Google Collab

pip install --quiet condacolab

Install dependencies

Important : PyTorch and Associated Libraries and CUDA Version

Python dependencies are installed using the environment.yaml file.

For PyTorch-CUDA, to avoid conflicts, do not install another version of PyTorch-CUDA if it is already installed.

Instead, make sure to remove any PyTorch-CUDA entries from the environment.yaml file, and to ensure compatibility with the CUDA version installed you need to install PyTorch, torchvision, and torchaudio using a specific CUDA version.

You can get the exact version with the command on the terminal of the system:

echo $CUDA_VERSION

Note : If you are using Docker or Singularity/Apptainer, you can use the inseefrlab/onyxia-jupyter-pytorch:py3.11.10-gpu image.

Initialyse Python environment

cd /home/onyxia/work/ (in order to install the new environment here - please change the directory to yours)

conda install -c conda-forge mamba

conda create -n 4dvarnet
# conda create -n 4dvarnet mamba python=3.9 -c conda-forge
conda activate 4dvarnet
mamba env update -y -f environment.yaml (please change the directory to yours)
mamba clean -a

If the PyTorch-CUDA was already installed in ypour environment, and you have removed y=them from the environment.yaml file, use this command to install PyTorch, pytorch-lightning, torchvision, and torchaudio libraries, ensuring they are built the right CUDA version support (12.2 in this exemple) :

mamba install -y pytorch  pytorch-lightning torchvision torchaudio pytorch-cuda=12.2 -c pytorch -c nvidia
mamba clean -a

If you get somme error messages, have a look at the "known-issues" documentation. In some situations, the version of numpy, or component used by pyTorch must be adapted.

You also may need to install ipython and omegaconfig :

mamba install ipykernel omegaconfig

Run JupyterLab

  1. Start Jupyterlab
jupyter lab --no-browser --ip 0.0.0.0
  1. Connect to the jupyter session : http://localhost:8888