{
  "modality": "create",
  "encoder": "openai/clip-vit-base-patch32",
  "encoder_dim": 768,
  "n_pillars": 5,
  "pillars": [
    "hook",
    "hold",
    "algorithmic_fit",
    "brand_lift",
    "overall"
  ],
  "d_model": 256,
  "n_layers": 4,
  "max_seq_len": 16,
  "vocab_size": 138,
  "bos_id": 1,
  "eos_id": 2,
  "pad_id": 0,
  "sep_id": 3
}