config.json

{
  "_name_or_path": "base-v1_sd",
  "architectures": [
    "OwlViTForObjectDetection"
  ],
  "initializer_factor": 1.0,
  "logit_scale_init_value": 2.6592,
  "model_type": "owlvit",
  "projection_dim": 512,
  "text_config": {
    "bos_token_id": 0,
    "dropout": 0.0,
    "eos_token_id": 2,
    "model_type": "owlvit_text_model",
    "pad_token_id": 1
  },
  "text_config_dict": null,
  "torch_dtype": "float32",
  "transformers_version": "4.45.1",
  "vision_config": {
    "dropout": 0.0,
    "max_length": 16,
    "model_type": "owlvit_vision_model",
    "patch_size": 16
  },
  "vision_config_dict": null
}