{
  "_name_or_path": "togethercomputer/Dragonfly-med-v1-llama8b",
  "architectures": [
    "DragonflyForCausalLM"
  ],
  "attention_dropout": 0.0,
  "bos_token_id": 128000,
  "eos_token_id": 128001,
  "hidden_act": "silu",
  "hidden_dropout": 0.0,
  "hidden_size": 4096,
  "image_encoder": "openai/clip-vit-base-patch32",
  "image_size": 300,
  "initializer_range": 0.02,
  "intermediate_size": 14336,
  "layer_norm_eps": 1e-05,
  "max_position_embeddings": 8192,
  "model_type": "dragonfly",
  "num_attention_heads": 32,
  "num_channels": 3,
  "num_hidden_layers": 32,
  "num_key_value_heads": 8,
  "partial_rotary_factor": 0.5,
  "patch_size": 30,
  "qk_layernorm": true,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 500000.0,
  "text_config": {
    "attention_bias": false,
    "bos_token_id": 128000,
    "eos_token_id": 128001,
    "hidden_act": "silu",
    "intermediate_size": 14336,
    "max_position_embeddings": 8192,
    "model_type": "llama",
    "num_attention_heads": 32,
    "num_hidden_layers": 32,
    "num_key_value_heads": 8,
    "pretraining_tp": 1,
    "rms_norm_eps": 1e-05,
    "rope_theta": 500000.0,
    "vocab_size": 128256
  },
  "text_pretrained_model_name_or_path": "meta-llama/Meta-Llama-3-8B-Instruct",
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.40.0",
  "use_cache": true,
  "use_imgconst_loss": false,
  "use_imgconst_loss_ve": false,
  "vocab_size": 128256
}