# core
core:
  version: "musicfm_7q"
  precision: "bf16-mixed" # "32-true" or "16-mixed" or "bf16-mixed"
  save_top_k: 2
  devices: 8
  num_nodes: 2
  max_epochs: 999999999
  ckpt_path: null

# data
data:
  batch_size: 21
  num_workers: 8
  limit_train: 1.0
  num_val_samples: -1

# optimization
optim:
  learning_rate: 1e-4

# model parameters
model:
  version: "7q"
  num_codebooks: 1
  codebook_dim: 16
  codebook_size: 4096
  num_latent_codebook: 4096
  latent_codebook_dim: 16
  features: ["melspec_2048"]
  hop_length: 240
  n_mels: 128
  conv_dim: 512
  encoder_dim: 1024
  encoder_depth: 12
  mask_hop: 0.4
  mask_prob: 0.6
  is_flash: True
  stat_path: "/app/suno/minz/models/mertlong_stats.json"
  model_path: null