{
  "sample_rate": 24000,
  "n_fft": 768,
  "hop_length": 240,
  "d_model": 512,
  "latent_dim": 128,
  "n_layers_per_stage": 4,
  "n_heads": 8,
  "dim_ff": 1024,
  "n_stages": 5,
  "context_window": 112,
  "causal": true,
  "activation": "swiglu",
  "center": false,
  "head_parametrisation": "complex"
}