:py:mod:`n2j.trainer`
=====================

.. py:module:: n2j.trainer

.. autoapi-nested-parse::

   Class managing the model training



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

Classes
~~~~~~~

.. autoapisummary::

   n2j.trainer.Trainer



Functions
~~~~~~~~~

.. autoapisummary::

   n2j.trainer.is_decreasing



.. py:function:: is_decreasing(arr)

   Returns True if array ever decreased

       


.. py:class:: Trainer(device_type, checkpoint_dir='trained_models', seed=123)

   .. py:method:: seed_everything()

      Seed the training and sampling for reproducibility

              


   .. py:method:: load_dataset(data_kwargs, is_train, batch_size, sub_features=None, sub_target=None, sub_target_local=None, rebin=False, num_workers=2, noise_kwargs={'mag': {'override_kwargs': None, 'depth': 5}}, detection_kwargs={})

      Load dataset and dataloader for training or validation

      Note
      ----
      Should be called for training data first, to set the normalizing
      stats used for both training and validation!



   .. py:method:: configure_model(model_name, model_kwargs={})


   .. py:method:: load_state(state_path)

      Load the state dict of the past training

      Parameters
      ----------
      state_path : str or osp.object
          path of the state dict to load



   .. py:method:: save_state(train_loss, val_loss)

      Save the state dict of the current training to disk

      Parameters
      ----------
      train_loss : float
          current training loss
      val_loss : float
          current validation loss



   .. py:method:: configure_optim(early_stop_memory=20, weight_local_loss=0.1, optim_kwargs={}, lr_scheduler_kwargs={'factor': 0.5, 'min_lr': 1e-07})

      Configure optimization-related objects

              


   .. py:method:: train_single_epoch(epoch_i)


   .. py:method:: train(n_epochs, sample_kwargs={})


   .. py:method:: infer(epoch_i)


   .. py:method:: _log_kappa_recovery_flow(epoch_i, x, u, y)


   .. py:method:: _log_kappa_recovery(epoch_i, x, u, y)


   .. py:method:: __repr__()

      Return repr(self).