rhիLdZddlZddlmZddlmZddlmZm Z m Z ddl Z ddl Z ddlZ ddl mZddlmZmZmZdd lmZdd lmZdd lmZmZdd lmZmZdd lmZm Z ddl!m"Z"m#Z#m$Z$ddl%m&Z&m'Z'ddl(m)Z)e$jTe+Z,ee#dGdde"Z-ee#dGdde"Z.dZ/Gddej`Z1Gddej`Z2 dAdej`de jfd e jfd!e jfd"e e jfd#e4d$e4fd%Z5Gd&d'ej`Z6Gd(d)ej`Z7Gd*d+ej`Z8Gd,d-ej`Z9Gd.d/ej`Z:Gd0d1eZ;Gd2d3ej`ZGd8d9ej`Z?e#d:Gd;dd?e=ZAgd@ZBy)Bz,PyTorch VideoMAE (masked autoencoder) model.N)deepcopy) dataclass)CallableOptionalUnion)nn)BCEWithLogitsLossCrossEntropyLossMSELoss)ACT2FN)GradientCheckpointingLayer)BaseModelOutputImageClassifierOutput)ALL_ATTENTION_FUNCTIONSPreTrainedModel) find_pruneable_heads_and_indicesprune_linear_layer) ModelOutputauto_docstringlogging)IMAGENET_DEFAULT_MEANIMAGENET_DEFAULT_STD)VideoMAEConfigz[ Class for VideoMAEDecoder's outputs, with potential hidden states and attentions. ) custom_introceZdZUdZdZeejed<dZ ee ejed<dZ ee ejed<y)VideoMAEDecoderOutputz logits (`torch.FloatTensor` of shape `(batch_size, patch_size ** 2 * num_channels)`): Pixel reconstruction logits. Nlogits hidden_states attentions) __name__ __module__ __qualname____doc__rrtorch FloatTensor__annotations__r tupler!/var/www/html/ai-insurance-compliance-backend/venv/lib/python3.12/site-packages/transformers/models/videomae/modeling_videomae.pyrr-sR +/FHU&& '.8