etc.yaml

tag: ${hydra.job.config_name}
log_dir_name: "${now:%Y-%m-%d}/${now:%H-%M-%S}"
log_dir_path: ./logs/${tag}/${log_dir_name}

hydra:
  run:
    dir: ${log_dir_path} 
  job_logging:
    handlers:
      file:
        filename: ${hydra.run.dir}/${hydra.job.name}_${now:%Y-%m-%d}_${now:%H-%M-%S}.log  

defaults:
  - override hydra/job_logging: colorlog
  - override hydra/hydra_logging: colorlog

NOTE_OCTAVES: [E2, F2, F#2, G2, G#2, A2, A#2, B2, C3, C#3, D3, D#3, E3, F3, F#3, G3, G#3, A3, A#3, B3, C4, C#4, D4, D#4, E4, F4, F#4, G4, G#4, A4, A#4, B4, C5, C#5, D5, D#5, E5, F5, F#5, G5, G#5, A5, A#5, B5, C6, C#6, D6]
CHROMATIC_SCALE: [C, C#, D, D#, E, F, F#, G, G#, A, A#, B]

num_strings: 6
log_level: "INFO"
dry_run: false
raise_error: true
comment: ""
random_seed: 1
device: "cuda"
use_amp: false
tqdm: true
tb_log_dir: tensorboard
log_interval_steps: 100
num_workers: 8
print_examples_validation: true
validate_interval_updates: 1000
validate_interval_epochs: 100
evaluate_before_training: false
class_probabilities: false
batch_size: 64
trainer: ClassifierTrainer
run_test_at_end: false
save_weights: true
segment_audio_frames: null

preprocessing:
  force_reprocess: False
  skip_empty_notes: True

audio:
  feature_extractor: raw      # Choose from torchaudio, librosa, raw
  loader: torchaudio          # Choose from torchaudio, librosa
  sr: 24000                   # Target sample rate
  mono: true                  # Set to true to set audio to mono
  resample: true              # Set to true to resample audio
  audio_load_sec: 5         # Duration of audio loaded in seconds. Null loads all the audio. 
  slide_window_sec: 2.5       # Duration of the sliding window in seconds. Only used if audio_load_sec is not null.
  normalize: false            # Set to true if audio should be normalized
  feature: cqt                # Choose from cqt, spectrogram, melspectrogram, chroma_stft, raw
  device: ${device}
  chroma_stft_params:
    sr: ${audio.sr}
    n_fft: 2048
    hop_length: 512
    n_chroma: 84
  cqt_params: 
    sr: ${audio.sr}
    n_bins: 84
    hop_length: 512
    bins_per_octave: 12
  spectrogram_params:
    n_fft: 512
    win_length: 64
    hop_length: 64
  melspectrogram_params:
    n_fft: 2048
    win_length: null
    hop_length: null
    n_mels: 80
    f_min: 0
    f_max: null

data:
  num_strings: ${num_strings}
  log_level: ${log_level}
  target_len: 100
  target_num_classes: 23
  target_num_strings: ${num_strings}
  features_dir: null
  segments_dir: "./data/segments"
  targets_dir: "./data/targets"
  timestep_change: "random"  # "random" or value
  max_input_dim: null
  cache_data_mem: False
  audio_dir: "./data/audio"
  num_workers: ${num_workers}
  train_batch_size: 8 
  valid_batch_size: 1 
  train_shuffle: False
  test_batch_size: 1 
  train_csv_file: null
  valid_csv_file: null
  test_csv_file: null
  dummy_csv_file: null
  onset_target_feature_size: null
  augment_onset: 0
  skip_cache: true
  audio_augmentation: 
    pitch_shift_steps: null 

lr_scheduler: 
  updates:
    warmup:
      init_lr: 0.00001
      num_warmup: 1000
      warmup_strategy: cos
    name: ExponentialLR 
    gamma: 0.998
    interval: 100
  epoch: null 
  #   name: ReduceLROnPlateau 
  #   metric: loss
  #   factor: 0.95
  #   patience: 1