config.json

{
  "architectures": [
    "PaGEModel"
  ],
  "dim": 256,
  "dino_feature_dropout": 0.1,
  "dinov3_drop_path_rate": 0.0,
  "dinov3_hidden_size": 768,
  "dinov3_intermediate_size": 3072,
  "dinov3_layer_norm_eps": 1e-05,
  "dinov3_layerscale_value": 1.0,
  "dinov3_num_attention_heads": 12,
  "dinov3_num_hidden_layers": 12,
  "dinov3_num_register_tokens": 4,
  "dinov3_patch_size": 16,
  "dinov3_use_gated_mlp": false,
  "drop_path": 0.1,
  "dtype": "float32",
  "head_in_size": [
    256,
    256
  ],
  "heatmap_out_size": [
    64,
    64
  ],
  "image_mean": [
    0.485,
    0.456,
    0.406
  ],
  "image_std": [
    0.229,
    0.224,
    0.225
  ],
  "inout": true,
  "mlp_layer": "geglu",
  "mlp_ratio": 4.0,
  "model_type": "page",
  "n_head_self_attn_layers": 1,
  "n_reg_tokens": 4,
  "n_scene_head_interaction_layers": 5,
  "n_scene_self_attn_layers": 1,
  "num_heads": 8,
  "pos_encoding": "rope",
  "scene_in_size": [
    512,
    512
  ],
  "transformers_version": "5.6.2",
  "use_head_prompt": false,
  "auto_map": {
    "AutoConfig": "Octopus1/PaGE--modeling_page.PaGEConfig",
    "AutoModel": "Octopus1/PaGE--modeling_page.PaGEModel"
  }
}