`i "dZddlZddlmZddlmZmZmZmZddl Z ddl m Z ddl m Z ddl mZdd lmZmZmZmZdd lmZmZdd lmZmZmZdd lmZmZmZmZm Z d dl!m"Z"m#Z#m$Z$e j%e&Z'eedGddeZ(eedGddeZ)eeGddeZ*de j+de j+fdZ,de j+de j+fdZ-de$de.fdZ/d[d!ee.e0fd"e1fd#Z2Gd$d%e j3Z4Gd&d'e j5Z6Gd(d)e j3Z7Gd*d+e j3Z8Gd,d-e j3Z9Gd.d/e j3Z:Gd0d1e j3Z;Gd2d3e j3Z<Gd4d5e j3Z= d\d7e j3d8e j+d9e j+d:e j+d;ee j+dd=e>d>ee j+fd?Z?Gd@dAe j3Z@GdBdCe j3ZAGdDdEe j3ZBGdFdGe j3ZCGdHdIe j3ZDGdJdKeZEGdLdMe j3ZFGdNdOe j3ZGeGdPdQeZHedRGdSdTeHZIedUGdVdWeHZJeGdXdYeHZKgdZZLdS)]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<dS)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#tuple|/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/transformers/models/align/modeling_align.pyr r )sl 15L(5,-44459x 129998