vikp/surya_layout3
1118k
1{2 "_name_or_path": "vikp/layout4",3 "architectures": [4 "EfficientViTForSemanticSegmentation"5 ],6 "attention_probs_dropout_prob": 0.0,7 "classifier_dropout_prob": 0.0,8 "decoder_hidden_size": 512,9 "decoder_layer_hidden_size": 128,10 "depths": [11 1,12 1,13 1,14 6,15 616 ],17 "head_dim": 32,18 "hidden_dropout_prob": 0.0,19 "hidden_sizes": [20 32,21 64,22 160,23 25624 ],25 "id2label": {26 "0": "Blank",27 "1": "Caption",28 "2": "Footnote",29 "3": "Formula",30 "4": "List-item",31 "5": "Page-footer",32 "6": "Page-header",33 "7": "Picture",34 "8": "Section-header",35 "9": "Table",36 "10": "Text",37 "11": "Title"38 },39 "initializer_range": 0.02,40 "label2id": {41 "Blank": 0,42 "Caption": 1,43 "Footnote": 2,44 "Formula": 3,45 "List-item": 4,46 "Page-footer": 5,47 "Page-header": 6,48 "Picture": 7,49 "Section-header": 8,50 "Table": 9,51 "Text": 10,52 "Title": 1153 },54 "layer_norm_eps": 1e-06,55 "model_type": "efficientvit",56 "num_channels": 3,57 "num_classes": 2,58 "num_stages": 4,59 "patch_size": [60 7,61 762 ],63 "semantic_loss_ignore_index": -1,64 "strides": [65 2,66 2,67 2,68 2,69 270 ],71 "torch_dtype": "float32",72 "transformers_version": "4.42.4",73 "widths": [74 32,75 64,76 128,77 256,78 51279 ]80}81 