config.json

{
  "act_fn": "quick_gelu",
  "alignment_head": 2,
  "alignment_layer": 68,
  "architectures": [
    "JukeboxPrior"
  ],
  "attention_multiplier": 0.25,
  "attention_pattern": "large_separated_enc_dec_w_lyrics",
  "attn_dropout": 0,
  "attn_res_scale": false,
  "blocks": 64,
  "conv_res_scale": null,
  "emb_dropout": 0,
  "encoder_config": {
    "act_fn": "quick_gelu",
    "alignment_head": 2,
    "alignment_layer": 68,
    "attention_multiplier": 0.25,
    "attention_pattern": "raw_column_previous_row_attention",
    "attn_dropout": 0.0,
    "attn_res_scale": false,
    "blocks": 32,
    "conv_res_scale": null,
    "emb_dropout": 0.0,
    "encoder_config": null,
    "encoder_loss_fraction": 0.4,
    "hidden_size": 1280,
    "init_scale": 0.1,
    "is_encoder_decoder": false,
    "level": 0,
    "lyric_vocab_size": 80,
    "mask": true,
    "max_duration": 600,
    "max_nb_genres": 1,
    "merged_decoder": false,
    "metadata_conditioning": true,
    "metadata_dims": [
      604,
      7898
    ],
    "min_duration": 0,
    "mlp_multiplier": 1.0,
    "model_type": "jukebox_prior",
    "music_vocab_size": 2048,
    "n_ctx": 6144,
    "n_heads": 4,
    "nb_relevant_lyric_tokens": 384,
    "num_layers": 18,
    "res_conv_depth": 3,
    "res_conv_width": 128,
    "res_convolution_multiplier": 1,
    "res_dilation_cycle": null,
    "res_dilation_growth_rate": 1,
    "res_downs_t": [
      3,
      2,
      2
    ],
    "res_strides_t": [
      2,
      2,
      2
    ],
    "resid_dropout": 0.0,
    "sampling_rate": 44100,
    "spread": null,
    "timing_dims": 64,
    "zero_out": false
  },
  "encoder_loss_fraction": 0.4,
  "hidden_size": 4800,
  "init_scale": 0.2,
  "is_encoder_decoder": false,
  "level": 0,
  "lyric_vocab_size": 80,
  "mask": true,
  "max_duration": 600.0,
  "max_nb_genres": 5,
  "merged_decoder": true,
  "metadata_conditioning": true,
  "metadata_dims": [
    120,
    4111
  ],
  "min_duration": 0,
  "mlp_multiplier": 1.0,
  "model_type": "jukebox_prior",
  "music_vocab_size": 2048,
  "n_ctx": 8192,
  "n_heads": 8,
  "nb_relevant_lyric_tokens": 512,
  "num_layers": 79,
  "res_conv_depth": null,
  "res_conv_width": null,
  "res_convolution_multiplier": null,
  "res_dilation_cycle": null,
  "res_dilation_growth_rate": null,
  "res_downs_t": [
    3,
    2,
    2
  ],
  "res_strides_t": [
    2,
    2,
    2
  ],
  "resid_dropout": 0,
  "sampling_rate": 44100,
  "spread": null,
  "timing_dims": 128,
  "torch_dtype": "float32",
  "zero_out": false
}