dacapo.experiments.starts.start
===============================

.. py:module:: dacapo.experiments.starts.start


Attributes
----------

.. autoapisummary::

   dacapo.experiments.starts.start.logger
   dacapo.experiments.starts.start.head_keys


Classes
-------

.. autoapisummary::

   dacapo.experiments.starts.start.Start


Functions
---------

.. autoapisummary::

   dacapo.experiments.starts.start.match_heads


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

.. py:data:: logger

.. py:data:: head_keys
   :value: ['prediction_head.weight', 'prediction_head.bias', 'chain.1.weight', 'chain.1.bias']


.. py:function:: match_heads(model, head_weights, old_head, new_head)

   Matches the head of the model to the new head by copying the weights
   of the old head to the new head. The weights of the old head are
   copied to the new head by matching the labels of the old head to the
   labels of the new head.

   :param model: obj
                 The model to which the weights are to be loaded.
   :param head_weights: dict
                        The weights of the old head.
   :param old_head: list
                    The labels of the old head.
   :param new_head: list
                    The labels of the new head.

   :returns:

             obj
                 The model with the weights of the old head copied to the new
                 head.
   :rtype: model

   :raises RuntimeError: If the old head is not found in the new head, a RuntimeError
       exception is thrown which is logged and handled by loading
       only the common layers from weights.

   .. rubric:: Examples

   >>> model = match_heads(model, head_weights, old_head, new_head)

   .. rubric:: Notes

   This function is called by the Start class to match the head of
   the model to the new head by copying the weights of the old head
   to the new head.


.. py:class:: Start(start_config)



   This class interfaces with the dacapo store to retrieve and load the
   weights of the starter model used for finetuning.

   .. attribute:: run

      str
      The specified run to retrieve weights for the model.

   .. attribute:: criterion

      str
      The policy that was used to decide when to store the weights.

   .. attribute:: channels

      int
      The number of channels in the input data.

   .. method:: __init__(start_config)

      
      Initializes the Start class with specified config to run the
      initialization of weights for a model associated with a specific
      criterion.

   .. method:: initialize_weights(model, new_head=None)

      
      Retrieves the weights from the dacapo store and load them into
      the model.

   .. rubric:: Notes

   This class is used to retrieve and load the weights of the starter
   model used for finetuning from the dacapo store.


   .. py:attribute:: channels
      :value: None



   .. py:method:: initialize_weights(model, new_head=None)

      Retrieves the weights from the dacapo store and load them into
      the model.

      :param model: obj
                    The model to which the weights are to be loaded.
      :param new_head: list
                       The labels of the new head.

      :returns:

                obj
                    The model with the weights loaded from the dacapo store.
      :rtype: model

      :raises RuntimeError: If weights of a non-existing or mismatched layer are being
          loaded, a RuntimeError exception is thrown which is logged
          and handled by loading only the common layers from weights.

      .. rubric:: Examples

      >>> model = start.initialize_weights(model, new_head)

      .. rubric:: Notes

      This function is called by the Start class to retrieve the weights
      from the dacapo store and load them into the model.