config.json

{
  "EE_config": {
    "alpha": 0.5,
    "apply_ocr": true,
    "batch_size": 4,
    "benchmark_OCR": false,
    "checkpoint": "",
    "data_parallel": false,
    "dataset": "maveriq/tobacco3482",
    "device": "cuda",
    "downsampling": 0,
    "encoder_layer_strategy": "ramp",
    "epochs": 100,
    "eval_batch_size": 1,
    "eval_start": false,
    "exit_head_num_layers": 2,
    "exit_threshold": -1,
    "exits": [
      "text_avg",
      "vision_avg",
      1,
      4,
      8
    ],
    "forward_signature": [
      "pixel_values",
      "head_mask",
      "labels",
      "output_attentions",
      "output_hidden_states",
      "return_dict"
    ],
    "gamma": 0,
    "get_raw_ocr_data": false,
    "global_threshold": 1.000001,
    "gradient_accumulation_steps": 16,
    "inference_strategy": "max_confidence",
    "lowercase": false,
    "lr": 2e-05,
    "model": "dit",
    "model_weights": "microsoft/dit-base",
    "optimizer": "AdamW",
    "plot_exits": false,
    "print_freq": 50,
    "seed": 42,
    "temperature": 1,
    "test_dataset": "jordyvl/rvl_cdip_100_examples_per_class",
    "training_strategy": "joint_weighted_avg",
    "use_images": true,
    "warmup_ratio": 0,
    "weight_decay": 0
  },
  "_name_or_path": "microsoft/dit-base",
  "architectures": [
    "BeitForImageClassification"
  ],
  "attention_probs_dropout_prob": 0.0,
  "auxiliary_channels": 256,
  "auxiliary_concat_input": false,
  "auxiliary_loss_weight": 0.4,
  "auxiliary_num_convs": 1,
  "drop_path_rate": 0.1,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 768,
  "id2label": {
    "0": "ADVE",
    "1": "Email",
    "2": "Form",
    "3": "Letter",
    "4": "Memo",
    "5": "News",
    "6": "Note",
    "7": "Report",
    "8": "Resume",
    "9": "Scientific"
  },
  "image_size": 224,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "ADVE": 0,
    "Email": 1,
    "Form": 2,
    "Letter": 3,
    "Memo": 4,
    "News": 5,
    "Note": 6,
    "Report": 7,
    "Resume": 8,
    "Scientific": 9
  },
  "layer_norm_eps": 1e-12,
  "layer_scale_init_value": 0.1,
  "model_type": "beit",
  "num_attention_heads": 12,
  "num_channels": 3,
  "num_hidden_layers": 12,
  "out_indices": [
    3,
    5,
    7,
    11
  ],
  "patch_size": 16,
  "pool_scales": [
    1,
    2,
    3,
    6
  ],
  "problem_type": "single_label_classification",
  "semantic_loss_ignore_index": 255,
  "torch_dtype": "float32",
  "transformers_version": "4.26.1",
  "use_absolute_position_embeddings": true,
  "use_auxiliary_head": true,
  "use_mask_token": true,
  "use_mean_pooling": true,
  "use_relative_position_bias": false,
  "use_shared_relative_position_bias": false,
  "vocab_size": 8192
}