config.json

{
  "model_name": "fuse-clip-titok",
  "text_config": {
    "context_length": 180,
    "vocab_size": 57601,
    "width": 512,
    "heads": 8,
    "layers": 12,
    "mlp_ratio": 4.0,
    "ls_init_value": null,
    "no_causal_mask": true,
    "pad_id": 0,
    "proj_bias": false,
    "output_tokens": false,
    "pool_type": "eot"
  },
  "mask_pad": true,
  "image_tokenizer": "yucornetto/tokenizer_titok_bl128_vq8k_imagenet",
  "use_eoi": false,
  "context_length": 180,
  "transformer_size": "base",
  "mlm_probability": 0.1,
  "rand_repl_probability": 0.0
}