:py:mod:`squid.predictor`
=========================

.. py:module:: squid.predictor


Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   squid.predictor.BasePredictor
   squid.predictor.ScalarPredictor
   squid.predictor.ProfilePredictor
   squid.predictor.BPNetPredictor



Functions
~~~~~~~~~

.. autoapisummary::

   squid.predictor.predict_in_batches
   squid.predictor.profile_sum
   squid.predictor.profile_pca



.. py:class:: BasePredictor


   
   Base class for running inference on in silico mutated sequences.
















   ..
       !! processed by numpydoc !!
   .. py:attribute:: save_dir

      

   .. py:method:: __call__(x, x_ref=None, save_window=None)
      :abstractmethod:



.. py:class:: ScalarPredictor(pred_fun, task_idx=0, batch_size=64, save_window=None, **kwargs)


   Bases: :py:obj:`BasePredictor`

   
   Module for handling scalar-based model predictions.

   :param pred_fun: Function for returning model predictions.
   :type pred_fun: built-in function
   :param task_idx: Task index corresponding to a specific output head.
   :type task_idx: int
   :param batch_size: The number of predictions per batch.
   :type batch_size: int

   :returns: Batch of scalar predictions corresponding to inputs.
   :rtype: torch.Tensor















   ..
       !! processed by numpydoc !!
   .. py:method:: __call__(x, x_ref=None, save_window=None)



.. py:class:: ProfilePredictor(pred_fun, task_idx=0, batch_size=64, reduce_fun=np.sum, save_dir=None, save_window=None, **kwargs)


   Bases: :py:obj:`BasePredictor`

   
   Module for handling profile-based model predictions.

   :param pred_fun: Function for returning model predictions.
   :type pred_fun: function
   :param task_idx: Task index corresponding to a specific output head.
   :type task_idx: int
   :param batch_size: The number of predictions per batch.
   :type batch_size: int
   :param reduce_fun: Function for reducing profile prediction to scalar.
   :type reduce_fun: function

   :returns: Batch of scalar predictions corresponding to inputs.
   :rtype: torch.Tensor















   ..
       !! processed by numpydoc !!
   .. py:method:: __call__(x, x_ref=None, save_window=None)



.. py:class:: BPNetPredictor(pred_fun, task_idx=0, batch_size=64, reduce_fun='wn', axis=1, save_dir=None, save_window=None, **kwargs)


   Bases: :py:obj:`BasePredictor`

   
   Module for handling BPNet (Kipoi) model predictions.

   :param pred_fun: Function for returning model predictions.
   :type pred_fun: function
   :param task_idx: Task index corresponding to a specific output head.
   :type task_idx: int
   :param batch_size: The number of predictions per batch.
   :type batch_size: int
   :param reduce_fun: Function for reducing profile prediction to scalar.
   :type reduce_fun: function

   :returns: Batch of scalar predictions corresponding to inputs.
   :rtype: torch.Tensor















   ..
       !! processed by numpydoc !!
   .. py:method:: __call__(x, x_ref=None, save_window=None)



.. py:function:: predict_in_batches(x, x_ref=None, model_pred_fun=None, batch_size=None, task_idx=None, save_window=None, **kwargs)

   
   Function to compute model predictions in batch mode.

   :param x: One-hot sequences (shape: (N, L, A)).
   :type x: numpy.ndarray
   :param x_ref: One-hot reference sequence (shape: (L, A)). Required if save_window is not None.
   :type x_ref: numpy.ndarray, optional
   :param model_pred_fun: Built-in function for accessing model inference on inputs.
   :type model_pred_fun: function
   :param batch_size: The number of predictions per batch of model inference.
   :type batch_size: int
   :param save_window: Window used for delimiting sequences that are exported in 'x_mut' array
   :type save_window: [int, int], optional

   :returns: Model predictions.
   :rtype: numpy.ndarray















   ..
       !! processed by numpydoc !!

.. py:function:: profile_sum(pred, save_dir=None)

   
   Function to transform predictions to scalars using summation.

   :param pred: Batch of profile-based predictions.
   :type pred: np.ndarray

   :returns: Batch of scalar predictions.
   :rtype: numpy.ndarray















   ..
       !! processed by numpydoc !!

.. py:function:: profile_pca(pred, save_dir=None)

   
   Function to transform predictions to scalars using principal component analysis (PCA).

   :param pred: Batch of profile-based predictions.
   :type pred: np.ndarray

   :returns: Batch of scalar predictions formed from projection of embedded
             profiles onto the first principal component.
   :rtype: numpy.ndarray















   ..
       !! processed by numpydoc !!