{
  "_name_or_path": "GroundingGPT-7B",
  "architectures": [
    "LEGOLlamaForCausalLM"
  ],
  "attention_bias": false,
  "bos_token_id": 1,
  "eos_token_id": 2,
  "freeze_mm_mlp_adapter": false,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "image_aspect_ratio": "pad",
  "image_grid_pinpoints": null,
  "image_token_len": 576,
  "initializer_range": 0.02,
  "intermediate_size": 11008,
  "max_frame": 64,
  "max_position_embeddings": 4096,
  "mm_use_im_patch_token": true,
  "mm_use_im_start_end": true,
  "mm_vision_hidden_size": 1024,
  "mm_vision_select_feature": "patch",
  "mm_vision_select_layer": -2,
  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
  "model_type": "LEGO",
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "num_key_value_heads": 32,
  "num_query_token": 32,
  "pad_token_id": 0,
  "pretraining_tp": 1,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 10000.0,
  "sound_end_token": 32008,
  "sound_patch_token": 32006,
  "sound_start_token": 32007,
  "sound_token_len": 8,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.34.0",
  "tune_mm_mlp_adapter": false,
  "use_cache": true,
  "use_mm_vision_proj": true,
  "video_end_token": 32005,
  "video_patch_token": 32003,
  "video_start_token": 32004,
  "video_token_len": 64,
  "vocab_size": 32009
}