lenskit.graphs.lightgcn.LightGCNTrainer
=======================================

.. py:class:: lenskit.graphs.lightgcn.LightGCNTrainer(scorer, data, options)
   

   Bases: :py:obj:`lenskit.training.ModelTrainer`


   Protocol implemented by iterative trainers for models.  Models that
   implement :class:`UsesTrainer` will return an object implementing this
   protocol from their :meth:`~UsesTrainer.create_trainer` method.

   ..

       This protocol only defines the core aspects of training a model. Trainers
       should also implement :class:`~lenskit.state.ParameterContainer` to allow
       training to be checkpointed and resumed.

       It is also a good idea for the trainer to be pickleable, but the parameter
       container interface is the primary mechanism for checkpointing.

   :Stability: Full


   .. py:attribute:: scorer
      :type:  LightGCNScorer


   .. py:attribute:: data
      :type:  lenskit.data.Dataset


   .. py:attribute:: options
      :type:  lenskit.training.TrainingOptions


   .. py:attribute:: log
      :type:  structlog.stdlib.BoundLogger


   .. py:attribute:: rng
      :type:  numpy.random.Generator


   .. py:attribute:: device
      :type:  str


   .. py:attribute:: model
      :type:  torch_geometric.nn.LightGCN


   .. py:attribute:: matrix
      :type:  lenskit.data.MatrixRelationshipSet


   .. py:attribute:: coo
      :type:  lenskit.data.matrix.COOStructure


   .. py:attribute:: user_base
      :type:  int


   .. py:attribute:: edges
      :type:  torch.Tensor


   .. py:attribute:: optimizer
      :type:  torch.optim.Optimizer


   .. py:attribute:: epochs_trained
      :type:  int
      :value: 0



   .. py:method:: train_epoch()

      Perform one epoch of the training process, optionally returning metrics
      on the training behavior.  After each training iteration, the mmodel
      must be usable.



   .. py:method:: finalize()

      Finish the training process, cleaning up any unneeded data structures
      and doing any finalization steps to the model.

      The default implementation does nothing.



   .. py:method:: batch_loss(mb_edges, scores)
      :abstractmethod: