config.json

{
  "_name_or_path": "/media/ms/DATA2/data/tuning_dataset_7/encoder",
  "architectures": [
    "DonutForImageClassification"
  ],
  "attention_probs_dropout_prob": 0.0,
  "depths": [
    2,
    2,
    14,
    2
  ],
  "drop_path_rate": 0.1,
  "embed_dim": 128,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 1024,
  "id2label": {
    "0": "1040",
    "1": "1040_sch_1",
    "2": "1040_sch_2",
    "3": "1040_sch_3",
    "4": "1040_sch_8812",
    "5": "1040_sch_a",
    "6": "1040_sch_b",
    "7": "1040_sch_c",
    "8": "1040_sch_d",
    "9": "1040_sch_e",
    "10": "1040_sch_se",
    "11": "1040nr",
    "12": "1040nr_sch_oi",
    "13": "form_1125",
    "14": "form_8949",
    "15": "form_8959",
    "16": "form_8960",
    "17": "form_8995",
    "18": "form_8995_sch_a",
    "19": "letter",
    "20": "other_misc",
    "21": "w2"
  },
  "image_size": [
    2560,
    1920
  ],
  "initializer_range": 0.02,
  "label2id": {
    "1040": 0,
    "1040_sch_1": 1,
    "1040_sch_2": 2,
    "1040_sch_3": 3,
    "1040_sch_8812": 4,
    "1040_sch_a": 5,
    "1040_sch_b": 6,
    "1040_sch_c": 7,
    "1040_sch_d": 8,
    "1040_sch_e": 9,
    "1040_sch_se": 10,
    "1040nr": 11,
    "1040nr_sch_oi": 12,
    "form_1125": 13,
    "form_8949": 14,
    "form_8959": 15,
    "form_8960": 16,
    "form_8995": 17,
    "form_8995_sch_a": 18,
    "letter": 19,
    "other_misc": 20,
    "w2": 21
  },
  "layer_norm_eps": 1e-05,
  "mlp_ratio": 4.0,
  "model_type": "donut-swin",
  "num_channels": 3,
  "num_heads": [
    4,
    8,
    16,
    32
  ],
  "num_layers": 4,
  "patch_size": 4,
  "path_norm": true,
  "qkv_bias": true,
  "torch_dtype": "float32",
  "transformers_version": "4.41.2",
  "use_absolute_embeddings": false,
  "window_size": 10
}