config.json

{
  "_name_or_path": "jukebox-5b-lyrics",
  "architectures": [
    "JukeboxModel"
  ],
  "cond_c_res": [
    0,
    1,
    1
  ],
  "cond_depth": [
    3,
    16,
    16
  ],
  "cond_dilation_cycle": [
    null,
    8,
    8
  ],
  "cond_dilation_growth_rate": [
    1,
    3,
    3
  ],
  "cond_downs_t": [
    3,
    2,
    2
  ],
  "cond_m_conv": 1,
  "cond_res_scale": [
    null,
    true,
    false
  ],
  "cond_strides_t": [
    2,
    2,
    2
  ],
  "cond_width": [
    128,
    1024,
    1024
  ],
  "cond_zero_out": false,
  "copy_input": false,
  "fp16_params": true,
  "hop_fraction": [
    0.125,
    0.5,
    0.5
  ],
  "init_std": 0.2,
  "lyric_conditioning": [
    true,
    false,
    false
  ],
  "max_duration": 600.0,
  "max_nb_genres": 5,
  "merged_decoder": [
    true,
    false,
    false
  ],
  "metadata_conditioning": true,
  "metadata_dims": [
    [
      120,
      4111
    ],
    [
      120,
      4111
    ],
    [
      120,
      4111
    ]
  ],
  "min_duration": 23.8,
  "model_type": "jukebox",
  "nb_priors": 3,
  "nb_relevant_lyric_tokens": [
    512,
    0,
    0
  ],
  "lyric_enc_attn_dropout": 0.0,
  "lyric_enc_attn_order": [
    2,
    0,
    0
  ],
  "lyric_enc_blocks": 32,
  "lyric_enc_depth": [
    18,
    3,
    3
  ],
  "lyric_enc_emb_dropout": 0.0,
  "lyric_enc_heads": 4,
  "lyric_enc_init_scale": [
    0.1,
    0.4,
    0.4
  ],
  "lyric_enc_loss_fraction": [
    0.4,
    0.0,
    0.0
  ],
  "lyric_enc_m_attn": 0.25,
  "lyric_enc_m_mlp": 1.0,
  "lyric_enc_n_vocab": 80,
  "lyric_enc_pos_init": false,
  "lyric_enc_res_scale": false,
  "lyric_enc_resid_dropout": 0.0,
  "lyric_enc_spread": null,
  "lyric_enc_width": [
    1280,
    128,
    128
  ],
  "lyric_enc_zero_out": false,
  "prior_alignment_head": [
    2,
    null,
    null
  ],
  "prior_alignment_layer": [
    68,
    null,
    null
  ],
  "prior_attn_dropout": 0,
  "prior_attn_order": [
    10,
    2,
    2
  ],
  "prior_blocks": 128,
  "prior_depth": [
    79,
    72,
    72
  ],
  "prior_emb_dropout": 0,
  "prior_init_scale": [
    0.2,
    1,
    1
  ],
  "prior_latent_dim": 2048,
  "prior_m_attn": 0.25,
  "prior_n_ctx": [
    8192,
    8192,
    8192
  ],
  "prior_n_heads": [
    8,
    1,
    1
  ],
  "prior_pos_init": false,
  "prior_res_scale": false,
  "prior_resid_dropout": 0,
  "prior_spread": null,
  "prior_width": [
    4800,
    1920,
    1920
  ],
  "prior_zero_out": false,
  "sample_length": 1058304,
  "sampling_rate": 44100,
  "single_enc_dec": [
    false,
    false,
    false
  ],
  "timing_dims": 128,
  "torch_dtype": "float32",
  "transformers_version": "4.22.0.dev0",
  "vqvae_codebook_dimension": 2048,
  "vqvae_commit": 0.02,
  "vqvae_conv_block_depth": 4,
  "vqvae_conv_block_width": 32,
  "vqvae_depth": 4,
  "vqvae_dilation_cycle": null,
  "vqvae_dilation_growth_rate": 3,
  "vqvae_downs_t": [
    3,
    2,
    2
  ],
  "vqvae_emmbedding_width": 64,
  "vqvae_levels": 3,
  "vqvae_lmu": 0.99,
  "vqvae_m_conv": 1,
  "vqvae_multipliers": [
    2,
    1,
    1
  ],
  "vqvae_music_tokens_shapes": [
    [
      8268
    ],
    [
      33072
    ],
    [
      132288
    ]
  ],
  "vqvae_reverse_decoder_dilation": 1,
  "vqvae_strides_t": [
    2,
    2,
    2
  ],
  "vqvae_width": 64
}