lmflow.models.vision_encoder.clip_encoder
=========================================

.. py:module:: lmflow.models.vision_encoder.clip_encoder


Classes
-------

.. autoapisummary::

   lmflow.models.vision_encoder.clip_encoder.CLIPVisionTower


Functions
---------

.. autoapisummary::

   lmflow.models.vision_encoder.clip_encoder.build_vision_tower


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

.. py:function:: build_vision_tower(vision_tower_cfg, **kwargs)

.. py:class:: CLIPVisionTower(vision_tower, args, delay_load=False)

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


   .. py:attribute:: is_loaded
      :value: False



   .. py:attribute:: vision_tower_name


   .. py:attribute:: select_layer


   .. py:attribute:: select_feature


   .. py:method:: load_model()


   .. py:method:: encode_images(images, language_projection)


   .. py:method:: feature_select(image_forward_outs)


   .. py:method:: forward(images)


   .. py:property:: dummy_feature


   .. py:property:: dtype


   .. py:property:: device


   .. py:property:: config


   .. py:property:: hidden_size


   .. py:property:: num_patches


   .. py:method:: prepare_inputs_labels_for_multimodal(input_ids, attention_mask, past_key_values, labels, images, language_projection=None, language_model=None, **kwargs)

      
      Copy from the LLAVA code base.
      Should be polished.
















      ..
          !! processed by numpydoc !!