config.json

{
  "architectures": [
    "CapriForConditionalGeneration"
  ],
  "model_type": "capri",
  "text_model_name_or_path": "Qwen/Qwen2.5-0.5B",
  "vision_model_name_or_path": "google/siglip2-base-patch16-224",
  "adapter_subdir": "text_adapter",
  "projector_type": "mlp",
  "projector_in_dim": 768,
  "projector_hidden_dim": 3072,
  "projector_out_dim": 896,
  "image_token": "<image>",
  "image_token_id": 151665,
  "prompt_prefix": "<image> Caption:",
  "max_length": 64,
  "load_vision_tower_by_default": false,
  "processor_class": "CapriProcessor",
  "auto_map": {
    "AutoConfig": "configuration_capri.CapriConfig",
    "AutoModel": "modeling_capri.CapriForConditionalGeneration",
    "AutoModelForCausalLM": "modeling_capri.CapriForConditionalGeneration"
  }
}