config.json

{
  "architectures": [
    "CLIPModel"
  ],
  "class_labels_mapping": {
    "Ghost > Studio > bottom > front": 12,
    "Ghost > Studio > top > front": 11,
    "Ghost > Studio > zoomed > front": 10,
    "Human > Indoor > half-body > back": 9,
    "Human > Indoor > half-body > front": 8,
    "Human > Studio > full-body > back": 7,
    "Human > Studio > full-body > front": 6,
    "Human > Studio > full-body > side": 5,
    "Human > Studio > half-body > back": 4,
    "Human > Studio > half-body > front": 3,
    "Human > Studio > half-body > side": 2,
    "Human > Studio > zoomed > front": 1,
    "Human > Studio > zoomed > side": 0
  },
  "initializer_factor": 1.0,
  "logit_scale_init_value": 2.6592,
  "model_type": "clip",
  "projection_dim": 512,
  "text_config": {
    "bos_token_id": 0,
    "dropout": 0.0,
    "eos_token_id": 2,
    "model_type": "clip_text_model"
  },
  "torch_dtype": "float32",
  "transformers_version": "4.41.2",
  "vision_config": {
    "dropout": 0.0,
    "model_type": "clip_vision_model"
  }
}