gnn_tracking.training.ml
========================

.. py:module:: gnn_tracking.training.ml

.. autoapi-nested-parse::

   Pytorch lightning module with training and validation step for the metric learning
   approach to graph construction.



Classes
-------

.. autoapisummary::

   gnn_tracking.training.ml.MLModule


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

.. py:class:: MLModule(*, loss_fct: gnn_tracking.metrics.losses.MultiLossFct, gc_scanner: gnn_tracking.graph_construction.k_scanner.GraphConstructionKNNScanner | None = None, **kwargs)

   Bases: :py:obj:`gnn_tracking.training.base.TrackingModule`


   Pytorch lightning module with training and validation step for the metric
   learning approach to graph construction.


   .. py:attribute:: loss_fct
      :type:  gnn_tracking.metrics.losses.metric_learning.GraphConstructionHingeEmbeddingLoss


   .. py:attribute:: gc_scanner


   .. py:method:: get_losses(out: dict[str, Any], data: torch_geometric.data.Data) -> tuple[torch.Tensor, dict[str, float]]


   .. py:method:: training_step(batch: torch_geometric.data.Data, batch_idx: int) -> torch.Tensor | None


   .. py:method:: validation_step(batch: torch_geometric.data.Data, batch_idx: int)


   .. py:method:: on_validation_epoch_end() -> None


   .. py:method:: highlight_metric(metric: str) -> bool