:py:mod:`anomalib.models.fastflow.torch_model`
==============================================

.. py:module:: anomalib.models.fastflow.torch_model

.. autoapi-nested-parse::

   FastFlow Torch Model Implementation.



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

Classes
~~~~~~~

.. autoapisummary::

   anomalib.models.fastflow.torch_model.FastflowModel



Functions
~~~~~~~~~

.. autoapisummary::

   anomalib.models.fastflow.torch_model.subnet_conv_func
   anomalib.models.fastflow.torch_model.create_fast_flow_block



.. py:function:: subnet_conv_func(kernel_size: int, hidden_ratio: float) -> Callable

   Subnet Convolutional Function.

   Callable class or function ``f``, called as ``f(channels_in, channels_out)`` and
       should return a torch.nn.Module.
       Predicts coupling coefficients :math:`s, t`.

   :param kernel_size: Kernel Size
   :type kernel_size: int
   :param hidden_ratio: Hidden ratio to compute number of hidden channels.
   :type hidden_ratio: float

   :returns: Sequential for the subnet constructor.
   :rtype: Callable


.. py:function:: create_fast_flow_block(input_dimensions: List[int], conv3x3_only: bool, hidden_ratio: float, flow_steps: int, clamp: float = 2.0) -> anomalib.models.components.freia.framework.SequenceINN

   Create NF Fast Flow Block.

   This is to create Normalizing Flow (NF) Fast Flow model block based on
   Figure 2 and Section 3.3 in the paper.

   :param input_dimensions: Input dimensions (Channel, Height, Width)
   :type input_dimensions: List[int]
   :param conv3x3_only: Boolean whether to use conv3x3 only or conv3x3 and conv1x1.
   :type conv3x3_only: bool
   :param hidden_ratio: Ratio for the hidden layer channels.
   :type hidden_ratio: float
   :param flow_steps: Flow steps.
   :type flow_steps: int
   :param clamp: Clamp. Defaults to 2.0.
   :type clamp: float, optional

   :returns: FastFlow Block.
   :rtype: SequenceINN


.. py:class:: FastflowModel(input_size: Tuple[int, int], backbone: str, flow_steps: int, conv3x3_only: bool = False, hidden_ratio: float = 1.0)

   Bases: :py:obj:`torch.nn.Module`

   FastFlow.

   Unsupervised Anomaly Detection and Localization via 2D Normalizing Flows.

   :param input_size: Model input size.
   :type input_size: Tuple[int, int]
   :param backbone: Backbone CNN network
   :type backbone: str
   :param flow_steps: Flow steps.
   :type flow_steps: int
   :param conv3x3_only: Use only conv3x3 in fast_flow model. Defaults to False.
   :type conv3x3_only: bool, optinoal
   :param hidden_ratio: Ratio to calculate hidden var channels. Defaults to 1.0.
   :type hidden_ratio: float, optional

   :raises ValueError: When the backbone is not supported.

   .. py:method:: forward(self, input_tensor: torch.Tensor) -> Union[Tuple[List[torch.Tensor], List[torch.Tensor]], torch.Tensor]

      Forward-Pass the input to the FastFlow Model.

      :param input_tensor: Input tensor.
      :type input_tensor: Tensor

      :returns:

                During training, return
                    (hidden_variables, log-of-the-jacobian-determinants).
                    During the validation/test, return the anomaly map.
      :rtype: Union[Tuple[Tensor, Tensor], Tensor]


   .. py:method:: _get_cnn_features(self, input_tensor: torch.Tensor) -> List[torch.Tensor]

      Get CNN-based features.

      :param input_tensor: Input Tensor.
      :type input_tensor: Tensor

      :returns: List of features.
      :rtype: List[Tensor]


   .. py:method:: _get_cait_features(self, input_tensor: torch.Tensor) -> List[torch.Tensor]

      Get Class-Attention-Image-Transformers (CaiT) features.

      :param input_tensor: Input Tensor.
      :type input_tensor: Tensor

      :returns: List of features.
      :rtype: List[Tensor]


   .. py:method:: _get_vit_features(self, input_tensor: torch.Tensor) -> List[torch.Tensor]

      Get Vision Transformers (ViT) features.

      :param input_tensor: Input Tensor.
      :type input_tensor: Tensor

      :returns: List of features.
      :rtype: List[Tensor]