{
  "vocab_size": 49154,
  "d_model": 1024,
  "n_layers": 24,
  "n_heads": 16,
  "n_kv_heads": 4,
  "block_size": 4096,
  "mlp_ratio": 2.6666666666666665,
  "rope_theta": 100000.0,
  "dropout": 0.0,
  "grad_checkpoint": true,
  "mixer": "hybrid",
  "attn_every": 4,
  "gla_chunk": 64
}