config.json

{
  "architectures": [
    "LlavaForConditionalGeneration"
  ],
  "hidden_size": 7168,
  "ignore_index": -100,
  "image_token_index": 64000,
  "is_yi_vl_derived_model": true,
  "model_type": "llava",
  "pad_token_id": 64001,
  "projector_hidden_act": "gelu",
  "text_config": {
    "architectures": [
      "LlamaForCausalLM"
    ],
    "hidden_size": 7168,
    "intermediate_size": 20480,
    "max_position_embeddings": 4096,
    "model_type": "llama",
    "num_attention_heads": 56,
    "num_hidden_layers": 60,
    "num_key_value_heads": 8,
    "rms_norm_eps": 1e-05,
    "rope_theta": 5000000.0,
    "torch_dtype": "bfloat16",
    "vocab_size": 64064
  },
  "torch_dtype": "bfloat16",
  "transformers_version": "4.40.1",
  "use_cache": true,
  "vision_config": {
    "architectures": [
      "CLIPVisionModel"
    ],
    "dropout": 0.0,
    "hidden_act": "gelu",
    "hidden_size": 1280,
    "image_size": 448,
    "intermediate_size": 5120,
    "model_type": "clip_vision_model",
    "num_attention_heads": 16,
    "num_hidden_layers": 32,
    "patch_size": 14,
    "projection_dim": 1024,
  },
  "vision_feature_layer": -2,
  "vision_feature_select_strategy": "default"
}