Setup
======


DeePCK is available for Python 3.8+ on Linux, macOS and Windows. 


Install with conda
------------------

1. Install ``numpy`` and visualization packages for basic usage.

.. code:: bash

   conda install numpy matplotlib seaborn

2. DeePCK depends on ``cantera``, so use ``cantera/label/dev`` channel to install ``cantera`` .

.. code:: bash
   
   conda install --channel cantera/label/dev cantera

3. The implementation of `Model Module` is based on the deep learning framework ``pytorch``, install ``pytorch`` CPU version  via ``conda``.
   
.. code:: bash

   conda install pytorch

3. Note:  if you prefer training the DNN model on GPU, you might need to install ``pytorch`` GPU version. 
   The GPU version requires compatible NVIDIA drivers to be installed already. 
   Therefore, check the CUDA driver version on your platform and see the installation guide on `Pytorch offical website <https://pytorch.org/get-started/locally/>`_. 
   For CUDA version 10.1, run the command for instance: 

.. code:: bash

   conda install pytorch=1.7 torchvision cudatoolkit=10.1 -c pytorch 

4. ``easydict`` allows to access dict values as attributes and then install it via ``conda`` :

.. code:: bash

   conda install -c conda-forge easydict 

or via ``pip`` :

.. code:: bash

   pip install easydict

5. (optional) If you are working on the HPC Clusters scheduled with `slurm system <https://docs.hpc.sjtu.edu.cn/job/slurm.html>`_, it is highly recommended 
   to install ``mpi4py`` for massive parallelization via ``conda`` (especielly for `Data Module`). Check `mpi4py documentation <https://mpi4py.readthedocs.io/en/stable/>`_ for more instruction.

.. code:: bash

   conda install -c conda-forge mpi4py openmpi







Install with docker
--------------------

To be continued.
 

.. toctree::
   :maxdepth: 2