config.json

{
  "model": {
    "dim": 1024,
    "n_layers": 16,
    "head_dim": 64,
    "seqlen_t": false,
    "huber_c": null,
    "input_dim": 32,
    "encoder_input_dim": 32,
    "encoder_output_dim": 32,
    "encoder_latent_downsample_factor": 1,
    "encoder_sliding_window": 65536,
    "sliding_window": 65536,
    "xattn_sliding_window": 65536,
    "max_seqlen": 256,
    "max_chans": 512,
    "model_dtype": "bf16",
    "stft_global_sigma": 0.1,
    "adaptive_loss_weighting": false,
    "num_fine_time_pts": 32,
    "rope_dim": 4,
    "rope_theta": 10000.0,
    "ape_dim": 0,
    "ape_theta": 10000.0,
    "ape_embedding_type": "scaled_dim_sinusoidal",
    "tok_idx_type": "{x,y,z,tc}",
    "dont_noise_chan_xyz": false,
    "zero_spatial": false,
    "dropout_vec_type": "zeros",
    "register_tok_idx": "mean_all"
  }
}