Mamba: differenze tra le versioni

Versione attuale delle 14:51, 12 giu 2024

To install Miniforge - Mamba

   su
   rm -rf /opt/anaconda3
   wget "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"
   bash Miniforge3-$(uname)-$(uname -m).sh
       when installing change the path to: /opt/miniforge3
       when asked if you want to add conda init say: YES
   restart terminal
   su
   conda config --set auto_activate_base false
   restart terminal
   su
   conda activate
   conda install -c conda-forge mamba
   mamba install jupyter jupyter-forward nb_conda_kernels jupyter-server-proxy dask-labextension 

Now lets install the environments:

   METEOCEAN
   mamba env create -f /vhe/nasmetocean/wavewatch/py_libs/meteocean.yml
       RUN WITH: /opt/miniforge3/envs/meteocean/bin/python

   PANGEO
   mamba env create -f /vhe/nasmetocean/wavewatch/py_libs/pangeo.yml
       RUN WITH: /opt/miniforge3/envs/pangeo/bin/python

   MACHINE LEARNING
   mamba env create -f /vhe/nasmetocean/wavewatch/py_libs/machine-learning.yml
       RUN WITH: /opt/miniforge3/envs/machine-learning/bin/python
 - Check that CUDA is installed:
 - nvcc --version 
 - If it says that nvcc doesn’t exists check under /usr/local/ if you have a cuda folder. If not CUDA is not installed. Follow the following guide. 
 [1. Introduction — Installation Guide for Linux 12.4 documentation](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html)

- [CUDA Toolkit 12.1 Downloads](https://developer.nvidia.com/cuda-downloads)

 - wget https://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda_12.4.0_550.54.14_linux.run
 - sh cuda_12.4.0_550.54.14_linux.run

   mamba activate machine-learning
   pip install falkon -f https://falkon.dibris.unige.it/torch-2.2.0_cu121.html

   exit su --> wavewatch user
   conda activate
       if error because conda does not exist /opt/miniforge/bin/conda init
   restart terminal
   conda config --set auto_activate_base false
   mamba init
   restart terminal
   test everything is working:
       conda activate
       mamba activate ENV

Install OceanParcels package

   su
   cd /opt
   git clone https://github.com/OceanParcels/parcels.git
   cd parcels
   mamba env create -f environment.yml
   mamba activate parcels
   pip install --no-build-isolation --no-deps -e .
        RUN WITH: /opt/miniforge3/envs/parcels/bin/python

Set up jupyter-forward environments. For each ENV:

   su
   conda activate
   mamba activate ENV
   python -m ipykernel install --name ENV --display-name ENV

Try tunnelling to another workstation:

   conda activate
   jupyter-forward wavewatch@ned.dicca.unige.it
       if jupyter is not found:
           go to ned and check /bin/bash -lc "which jupyter"
           conda activate
           which python
           copy the path
           open .bashrc
           export PATH=/opt/miniforge3/bin/:$PATH

To update environments:

   su and add the packages to the corresponding ENV.yml (/vhe/nasmetocean/wavewatch/py_libs/)
   su
   conda activate
   mama activate ENV
   mamba env update --file /vhe/nasmetocean/wavewatch/py_libs/ENV.yml --prune