config.json

{
  "architectures": [
    "TinyLlavaForConditionalGeneration"
  ],
  "cache_dir": null,
  "connector_type": "mlp2x_gelu",
  "hidden_size": 1280,
  "ignore_index": -100,
  "image_aspect_ratio": "square",
  "image_token_index": -200,
  "llm_model_name_or_path": "apple/OpenELM-270M-Instruct",
  "model_type": "tinyllava",
  "num_queries": 128,
  "num_resampler_layers": 3,
  "pad_token": "<unk>",
  "pad_token_id": 0,
  "resampler_hidden_size": 768,
  "text_config": {
    "_name_or_path": "apple/OpenELM-270M-Instruct",
    "activation_fn_name": "swish",
    "architectures": [
      "OpenELMForCausalLM"
    ],
    "auto_map": {
      "AutoConfig": "apple/OpenELM-270M-Instruct--configuration_openelm.OpenELMConfig",
      "AutoModelForCausalLM": "apple/OpenELM-270M-Instruct--modeling_openelm.OpenELMForCausalLM"
    },
    "ffn_dim_divisor": 256,
    "ffn_multipliers": [
      0.5,
      0.73,
      0.97,
      1.2,
      1.43,
      1.67,
      1.9,
      2.13,
      2.37,
      2.6,
      2.83,
      3.07,
      3.3,
      3.53,
      3.77,
      4.0
    ],
    "ffn_with_glu": true,
    "head_dim": 64,
    "max_context_length": 2048,
    "model_dim": 1280,
    "model_type": "openelm",
    "normalization_layer_name": "rms_norm",
    "normalize_qk_projections": true,
    "num_gqa_groups": 4,
    "num_kv_heads": [
      3,
      3,
      3,
      3,
      3,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      5,
      5,
      5,
      5
    ],
    "num_query_heads": [
      12,
      12,
      12,
      12,
      12,
      16,
      16,
      16,
      16,
      16,
      16,
      16,
      20,
      20,
      20,
      20
    ],
    "num_transformer_layers": 16,
    "qkv_multipliers": [
      0.5,
      1.0
    ],
    "rope_freq_constant": 10000,
    "rope_max_length": 4096,
    "share_input_output_layers": true,
    "tie_word_embeddings": true,
    "torch_dtype": "float16"
  },
  "tokenizer_model_max_length": 2048,
  "tokenizer_name_or_path": "meta-llama/Llama-2-7b-hf",
  "tokenizer_padding_side": "right",
  "tokenizer_use_fast": false,
  "torch_dtype": "float16",
  "transformers_version": "4.39.3",
  "tune_type_connector": "full",
  "tune_type_llm": "full",
  "tune_type_vision_tower": "frozen",
  "tune_vision_tower_from_layer": 0,
  "use_cache": false,
  "vision_config": {
    "_name_or_path": "facebook/dinov2-small",
    "apply_layernorm": true,
    "architectures": [
      "Dinov2Model"
    ],
    "attention_probs_dropout_prob": 0.0,
    "drop_path_rate": 0.0,
    "hidden_act": "gelu",
    "hidden_dropout_prob": 0.0,
    "hidden_size": 384,
    "image_size": 518,
    "layer_norm_eps": 1e-06,
    "layerscale_value": 1.0,
    "mlp_ratio": 4,
    "model_name_or_path": "facebook/dinov2-small",
    "model_name_or_path2": "",
    "model_type": "dinov2",
    "num_attention_heads": 6,
    "num_hidden_layers": 12,
    "out_features": [
      "stage12"
    ],
    "out_indices": [
      12
    ],
    "patch_size": 14,
    "qkv_bias": true,
    "reshape_hidden_states": true,
    "stage_names": [
      "stem",
      "stage1",
      "stage2",
      "stage3",
      "stage4",
      "stage5",
      "stage6",
      "stage7",
      "stage8",
      "stage9",
      "stage10",
      "stage11",
      "stage12"
    ],
    "torch_dtype": "float32",
    "use_swiglu_ffn": false
  },
  "vision_feature_layer": -2,
  "vision_feature_select_strategy": "patch",
  "vision_hidden_size": 384,
  "vision_model_name_or_path": "facebook/dinov2-small",
  "vision_model_name_or_path2": "",
  "vocab_size": 32000
}