clap_com/runtime_config.json

{
  "CSV_PATH": "/mnt/bn/musicevalbigai/wav44100/dataset_raw/MADB_mean/song_avg_scores.csv",
  "EMB_DIR": "/mnt/bn/musicevalbigai/extractor/clap/output/musiceval_pretrained/12/",
  "SAVE_DIR": "/mnt/bn/musicevalbigai/nips2026/train/result/clap/7_s2_12/",
  "SHARED_INIT_PT": "",
  "TARGET_COLUMNS": [
    "avg_overall_score",
    "avg_melody_perception",
    "avg_melody_emotion",
    "avg_rhythm_perception",
    "avg_structure_perception",
    "avg_performance_and_singing_mood",
    "avg_performance_skill"
  ],
  "ID_COL": "id",
  "DEVICE": "cuda",
  "SEED": 5,
  "BATCH_SIZE": 64,
  "NUM_WORKERS": 2,
  "EPOCHS": 200,
  "LR": 0.0001,
  "WEIGHT_DECAY": 0.0,
  "VALID_RATIO": 0.2,
  "USE_AMP": false,
  "MODEL_KWARGS": {
    "input_dim": 512,
    "hidden_dims": 64,
    "mlp_hidden_dims": [
      64,
      16
    ],
    "seq_len": 8,
    "nhead": 8,
    "num_layers": 4,
    "dim_feedforward": 256,
    "dropout": 0.1,
    "out_dim": 1
  },
  "loss": {
    "use_mask": false,
    "use_pearson_loss": true,
    "pearson_loss_weight": 0.5,
    "train_formula": "mse + pearson_weight * (1 - pearson)"
  },
  "selection_metric": {
    "type": "per_dimension_best_only",
    "composite_formula": {
      "mse": 1.0,
      "1_minus_pearson": 0.33,
      "1_minus_spearman": 0.33,
      "1_minus_kendall": 0.33
    }
  }
}