developer.rst

Developer guide

Here we feature tutorials to guide you through the construction of a model with scvi-tools. For an example of how scvi-tools can be used in an independent package, see our GitHub template.

Note

For questions about developing with scvi-tools, please use our forum or gitter.

Model development tutorials

.. nbgallery::

   notebooks/data_tutorial
   notebooks/module_user_guide
   notebooks/model_user_guide

Glossary

.. tabs::

    .. tab:: Model

        A model class is a user-facing object that contains the module as an attribute (i.e., ``self.module``).
        The model has a `train` method that learns the parameters of the module, and also contains methods
        for users to retrieve information from the module, like the latent representation of cells in a VAE.
        Conventionally, the post-inference model methods should not store data into the AnnData object, but
        instead return "standard" Python objects, like numpy arrays or pandas dataframes.

    .. tab:: Module

        A module is the lower-level object that defines a generative model and inference scheme. A module will
        either inherit :class:`~scvi.module.base.BaseModuleClass` or :class:`~scvi.module.base.PyroBaseModuleClass`.
        Consequently, a module can either be implemented with PyTorch alone, or Pyro. In the PyTorch only case, the
        generative process and inference scheme are implemented respectively in the `generative` and `inference` methods,
        while the `loss` method computes the, e.g, ELBO in the case of variational inference.

.. tabs::

    .. tab:: TrainingPlan


        The training plan is a PyTorch Lightning Module that is initializd with a scvi-tools module object.
        It configures the optimizers, defines the training step and validation step, and computes metrics to be
        recorded during training. The training step and validation step are functions that take data, run it through
        the model and return the loss, which will then be used to optimize the model parameters in the Trainer.
        Overall, training plans can be used to develop complex inference schemes on top of modules.


    .. tab:: Trainer

        The :class:`~scvi.train.Trainer` is a lightweight wrapper of the PyTorch Lightning Trainer. It takes as input
        the training plan, a training data loader, and a validation dataloader. It performs the actual training loop, in
        which parameters are optimized, as well as the validation loop to monitor metrics. It automatically handles moving
        data to the correct device (CPU/GPU).