config.json

{
  "_name_or_path": "PLAN-Lab/CALICO",
  "architectures": [
    "CALICOForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "auto_initialize_adaptors": true,
  "auto_initialize_qformer": true,
  "bbox_token_idx": 32002,
  "bos_token_id": 1,
  "dino_hidden_size": 768,
  "dino_model_name": "dinov2_vitb14_reg",
  "eos_token_id": 2,
  "global_image_size": 224,
  "grounding_encoder": "sam_vit_h",
  "grounding_image_size": 1024,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "image_aspect": "square",
  "image_aspect_ratio": "square",
  "initializer_range": 0.02,
  "intermediate_size": 11008,
  "layer_type": "linear",
  "max_position_embeddings": 4096,
  "mlp_bias": false,
  "mm_projector_type": "linear",
  "model_type": "calico",
  "num_attention_heads": 32,
  "num_attn_heads": 8,
  "num_hidden_layers": 32,
  "num_key_value_heads": 32,
  "num_level_reg_features": 4,
  "num_query_tokens": 32,
  "out_dim": 256,
  "pad_token_id": 0,
  "pretraining_tp": 1,
  "q_former_model": "",
  "qformer_hidden_size": 768,
  "qformer_vision_encoder": "eva_clip_g",
  "qformer_vision_width": 1408,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 10000.0,
  "seg_image_tokens": [
    [
      32004,
      313,
      2382,
      29896
    ],
    [
      32004,
      313,
      2382,
      29906
    ]
  ],
  "seg_token_idx": 32004,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "train_mask_decoder": true,
  "transformers_version": "4.42.3",
  "update_layers": [
    11,
    22
  ],
  "use_cache": false,
  "use_mm_proj": true,
  "vocab_size": 32007,
  "with_region": true
}