Ph 2dZddlZddlmZddlmZmZmZmZddl Z ddl m Z ddl m Z ddl mZdd lmZmZmZmZdd lmZmZdd lmZmZmZdd lmZmZmZmZm Z d dl!m"Z"m#Z#m$Z$e jJe&Z'eedGddeZ(eedGddeZ)eeGddeZ*de jVde jVfdZ,de jVde jVfdZ-de$de.fdZ/dYd ee.e0fd!e1fd"Z2Gd#d$e jfZ4Gd%d&e jjZ6Gd'd(e jfZ7Gd)d*e jfZ8Gd+d,e jfZ9Gd-d.e jfZ:Gd/d0e jfZ;Gd1d2e jfZ<Gd3d4e jfZ= dZd5e jfd6e jVd7e jVd8e jVd9ee jVd:e>d;e>dd?e jfZ@Gd@dAe jfZAGdBdCe jfZBGdDdEe jfZCGdFdGe jfZDGdHdIeZEGdJdKe jfZFGdLdMe jfZGeGdNdOeZHedPGdQdReHZIedSGdTdUeHZJeGdVdWeHZKgdXZLy)[zPyTorch ALIGN model.N) dataclass)AnyCallableOptionalUnion)nn)ACT2FN)GradientCheckpointingLayer)BaseModelOutputBaseModelOutputWithNoAttentionBaseModelOutputWithPooling(BaseModelOutputWithPoolingAndNoAttention)ALL_ATTENTION_FUNCTIONSPreTrainedModel)apply_chunking_to_forward find_pruneable_heads_and_indicesprune_linear_layer) ModelOutputauto_docstringcan_return_tuplefilter_out_non_signature_kwargslogging) AlignConfigAlignTextConfigAlignVisionConfigz} Base class for vision model's outputs that also contains image embeddings of the pooling of the last hidden states. ) custom_introceZdZUdZdZeejed<dZ eejed<dZ ee ejed<y)AlignVisionModelOutputz image_embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)` *optional* returned when model is initialized with `with_projection=True`): The image embeddings obtained by applying the projection layer to the pooler_output. N image_embedslast_hidden_state hidden_states) __name__ __module__ __qualname____doc__r!rtorch FloatTensor__annotations__r"r#tupleb/var/www/html/saasai/venv/lib/python3.12/site-packages/transformers/models/align/modeling_align.pyr r )sN 15L(5,,-459x 1 1298