pyiqa.models.base_model
=======================

.. py:module:: pyiqa.models.base_model




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

.. py:class:: BaseModel(opt)

   Base model.


   .. py:method:: feed_data(data)


   .. py:method:: optimize_parameters()


   .. py:method:: get_current_visuals()


   .. py:method:: save(epoch, current_iter)

      Save networks and training state.



   .. py:method:: validation(dataloader, current_iter, tb_logger, save_img=False)

      Validation function.

      :param dataloader: Validation dataloader.
      :type dataloader: torch.utils.data.DataLoader
      :param current_iter: Current iteration.
      :type current_iter: int
      :param tb_logger: Tensorboard logger.
      :type tb_logger: tensorboard logger
      :param save_img: Whether to save images. Default: False.
      :type save_img: bool



   .. py:method:: model_ema(decay=0.999)


   .. py:method:: copy_model(net_a, net_b)

      copy model from net_a to net_b



   .. py:method:: get_current_log()


   .. py:method:: model_to_device(net)

      Model to device. It also warps models with DistributedDataParallel
      or DataParallel.

      :param net:
      :type net: nn.Module



   .. py:method:: get_optimizer(optim_type, params, lr, **kwargs)


   .. py:method:: setup_schedulers(scheduler_name='scheduler')

      Set up schedulers.



   .. py:method:: get_bare_model(net)

      Get bare model, especially under wrapping with
      DistributedDataParallel or DataParallel.



   .. py:method:: print_network(net)

      Print the str and parameter number of a network.

      :param net:
      :type net: nn.Module



   .. py:method:: update_learning_rate(current_iter, warmup_iter=-1)

      Update learning rate.

      :param current_iter: Current iteration.
      :type current_iter: int
      :param warmup_iter: Default： -1.
      :type warmup_iter: int



   .. py:method:: get_current_learning_rate()


   .. py:method:: save_network(net, net_label, current_iter=None, param_key='params')

      Save networks.

      :param net: Network(s) to be saved.
      :type net: nn.Module | list[nn.Module]
      :param net_label: Network label.
      :type net_label: str
      :param current_iter: Current iter number.
      :type current_iter: int
      :param param_key: The parameter key(s) to save network.
                        Default: 'params'.
      :type param_key: str | list[str]



   .. py:method:: load_network(net, load_path, strict=True, param_key='params')

      Load network.

      :param load_path: The path of networks to be loaded.
      :type load_path: str
      :param net: Network.
      :type net: nn.Module
      :param strict: Whether strictly loaded.
      :type strict: bool
      :param param_key: The parameter key of loaded network. If set to
                        None, use the root 'path'.
                        Default: 'params'.
      :type param_key: str



   .. py:method:: save_training_state(epoch, current_iter)

      Save training states during training, which will be used for
      resuming.

      :param epoch: Current epoch.
      :type epoch: int
      :param current_iter: Current iteration.
      :type current_iter: int



   .. py:method:: resume_training(resume_state)

      Reload the optimizers and schedulers for resumed training.

      :param resume_state: Resume state.
      :type resume_state: dict



   .. py:method:: reduce_loss_dict(loss_dict)

      reduce loss dict.

      In distributed training, it averages the losses among different GPUs .

      :param loss_dict: Loss dict.
      :type loss_dict: OrderedDict