emmi_inference.models.upt
=========================

.. py:module:: emmi_inference.models.upt

.. autoapi-nested-parse::

   WARNING
   This file is a 1:1 duplicate from the ``tutorial`` folder.

   It is here to avoid installation of a tutorial as a package and keep it isolated.



Classes
-------

.. autoapisummary::

   emmi_inference.models.upt.UPT


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

.. py:class:: UPT(ndim, radius, dim, num_heads, depth, dec_depth, mlp_expansion_factor, norm_ctor = nn.LayerNorm, spool_pos_mode = 'abspos', use_physics_features = False, physics_dim = None, **kwargs)

   Bases: :py:obj:`emmi_inference.models.base_model.BaseModel`


   Implementation of the UPT (Universal Physics Transformer) model.

   :param BaseModel: Base model class that contains the utilities for all models we use in this tutorial.

   :param ndim: rank of the input data/tensor.
   :param input_dim: dimensionality of the input features.
   :param output_dim: dimensionality of the output features.
   :param radius: radius for the supernode pooling.
   :param dim: hidden dimensionality of the network.
   :param num_heads: number of attention heads.
   :param depth: depth (i.e., number of layers) of the encoder/approximator.
   :param dec_depth: number of decoder layers for the perceiver decoder.
   :param mlp_expansion_factor: expansion factor for the MLP in the transformer block.
   :param norm_ctor: cons. Defaults to nn.LayerNorm.
   :param spool_pos_mode: _description_. Defaults to "abspos".


   .. py:attribute:: encoder


   .. py:attribute:: blocks


   .. py:attribute:: pred


   .. py:method:: forward(surface_mask_query, surface_position_batch_idx, surface_position_supernode_idx, surface_position, query_position, surface_features = None)

      Forward pass of the UPT model.

      :param input_position: input coordinates of the surface points.
      :param surface_mask_query: surface mask for the query points, indicating which points are surface points.
      :param surface_position_batch_idx: batch indices for the surface positions, since the surface positions are a sparse tensor for the supernode pooling.
      :param surface_position_supernode_idx: supernode indices for the surface positions.
      :param geometry_position: geometry position information.
      :param query_position: input coordinates of the query points.
      :param surface_features: surface features for the input points. Defaults to None.

      :returns: dictionary with the output tensors, containing the surface pressure and volume velocity.
      :rtype: dict[str, torch.Tensor]