popgym.baselines.models.embeddings
==================================

.. py:module:: popgym.baselines.models.embeddings


Classes
-------

.. autoapisummary::

   popgym.baselines.models.embeddings.PhaserEncoding
   popgym.baselines.models.embeddings.ScaledPositionalEncoding
   popgym.baselines.models.embeddings.SimpleScaledEncoding
   popgym.baselines.models.embeddings.Time2Vec
   popgym.baselines.models.embeddings.Base2Embedding
   popgym.baselines.models.embeddings.PerceptronEmbedding
   popgym.baselines.models.embeddings.SoftExponentialEmbedding


Functions
---------

.. autoapisummary::

   popgym.baselines.models.embeddings.positional_encoding


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

.. py:function:: positional_encoding(max_len, d_model)

.. py:class:: PhaserEncoding(max_len: int, d_model: int, num_dims: int, max_period: int = 1024, dtype: torch.dtype = torch.double, decay_bias: float = -0.5, freq_scale: float = 1.0, fudge_factor: float = 0.025, soft_clamp: bool = False)

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


   .. py:attribute:: num_dims


   .. py:attribute:: d_model


   .. py:attribute:: dtype


   .. py:attribute:: limit


   .. py:attribute:: soft_clamp
      :value: False



   .. py:attribute:: c


   .. py:attribute:: d


   .. py:attribute:: ab


   .. py:method:: psi(t_minus_i)


   .. py:method:: recurrent_update2(x, state)


   .. py:method:: recurrent_update(x, state)


   .. py:method:: forward(x, state)


.. py:class:: ScaledPositionalEncoding(num_embeddings, embedding_dim, **kwargs)

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


   .. py:attribute:: a


   .. py:method:: forward(idx)


.. py:class:: SimpleScaledEncoding(max_len, d_model)

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


   .. py:method:: forward(idx)


.. py:class:: Time2Vec(max_len, d_model)

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


   .. py:attribute:: d_model


   .. py:attribute:: max_len


   .. py:attribute:: net


   .. py:method:: forward(idx)


.. py:class:: Base2Embedding(max_len, d_model)

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


   .. py:attribute:: d_model


   .. py:attribute:: max_len


   .. py:method:: forward(idx)


.. py:class:: PerceptronEmbedding(max_len, d_model)

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


   .. py:attribute:: d_model


   .. py:attribute:: lin


   .. py:method:: forward(idx)


.. py:class:: SoftExponentialEmbedding(max_len, d_model)

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


   .. py:attribute:: d_model


   .. py:attribute:: a


   .. py:attribute:: lin


   .. py:method:: forward(idx)