dicee.models.flock_walks
========================

.. py:module:: dicee.models.flock_walks

.. autoapi-nested-parse::

   Flock's non-backtracking walk and recording protocol in PyTorch.

   Semantics follow jw9730/flock at f35103d25a78bdf4075de5c673a51de4979aa4d7.
   The RNG is PyTorch's CPU generator, not the reference C++ per-walk MT19937.
   Walk records can be supplied explicitly to compare neural computations exactly.



Classes
-------

.. autoapisummary::

   dicee.models.flock_walks.WalkGraph


Functions
---------

.. autoapisummary::

   dicee.models.flock_walks.compiled_walk_nodes
   dicee.models.flock_walks.anonymize


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

.. py:function:: compiled_walk_nodes()

.. py:function:: anonymize(values, missing_id=None, missing_name=None)

   Assign 1-based names in order of discovery, independently in each row.


.. py:class:: WalkGraph

   CPU CSR context, kept outside a model's transferable parameters.


   .. py:attribute:: edge_index
      :type:  torch.Tensor


   .. py:attribute:: edge_type
      :type:  torch.Tensor


   .. py:attribute:: num_nodes
      :type:  int


   .. py:attribute:: num_types
      :type:  int


   .. py:attribute:: compile_sampler
      :type:  bool
      :value: True



   .. py:method:: __post_init__()


   .. py:method:: walk(prefix, length, remove_loops, generator=None, prefix_types=None)

      Sample uniform neighbors, then a uniform typed edge in either direction.



   .. py:method:: parse_types(walks, generator=None)


   .. py:method:: sample(heads, tails, walk_num, length, refinements, generator=None)

      Return seven official-format records of shape [T,B,3N or 4N,L].