{
  "axes": [
    {
      "rows": 1536,
      "freqs": 16
    },
    {
      "rows": 512,
      "freqs": 24
    },
    {
      "rows": 512,
      "freqs": 24
    }
  ],
  "seq_multi_of": 32,
  "latent_scaling": 0.3611,
  "latent_shift": 0.1159,
  "note": "caption token i -> coord (i+1,0,0); image token (h,w) -> coord (n_cap+1,h,w). Concat the three axis lookups for cos and sin."
}