splitctc.yaml

# lightning.pytorch==1.9.5
seed_everything: true
trainer:
  logger: true
  enable_checkpointing: true
  callbacks: 
    - class_path: lightning.pytorch.callbacks.ModelCheckpoint
      init_args:
        dirpath: null
        filename: '{epoch}'
        monitor: V_SymbER
        verbose: false
        save_last: true
        save_top_k: 3
        save_weights_only: false
        mode: min
        auto_insert_metric_name: true
        every_n_train_steps: null
        train_time_interval: null
        every_n_epochs: null
        save_on_train_epoch_end: null
  default_root_dir: outputs/SplitCTC
  gradient_clip_val: null
  gradient_clip_algorithm: null
  num_nodes: 1
  num_processes: null
  devices: 1
  gpus: 1
  auto_select_gpus: null
  tpu_cores: null
  ipus: null
  enable_progress_bar: true
  overfit_batches: 0.0
  track_grad_norm: -1
  check_val_every_n_epoch: 1
  fast_dev_run: false
  accumulate_grad_batches: null
  max_epochs: 10000
  min_epochs: 50
  max_steps: -1
  min_steps: null
  max_time: '{''days'': 2}'
  limit_train_batches: null
  limit_val_batches: null
  limit_test_batches: null
  limit_predict_batches: null
  val_check_interval: null
  log_every_n_steps: 50
  accelerator: gpu
  strategy: null
  sync_batchnorm: false
  precision: 32
  enable_model_summary: true
  num_sanity_val_steps: 2
  resume_from_checkpoint: null
  profiler: null
  benchmark: null
  deterministic: null
  reload_dataloaders_every_n_epochs: 0
  auto_lr_find: false
  replace_sampler_ddp: true
  detect_anomaly: false
  auto_scale_batch_size: false
  plugins: null
  amp_backend: null
  amp_level: null
  move_metrics_to_cpu: false
  multiple_trainloader_mode: max_size_cycle
  inference_mode: true

# model config
model:
  encoder: 
    class_path: model.basic.CNNEncoder.CNNEncoder
  decoder: 
    class_path: model.basic.RNNDecoder.RNNDecoder
  ctc_utils: 
    class_path: utils.SplitCTCUtils.SplitCTCUtil
    init_args:
      lift_path: dataset/tokenizer/tokenizer_lift.json
      note_path: dataset/tokenizer/tokenizer_note.json
      pitch_path: dataset/tokenizer/tokenizer_pitch.json
      rhythm_path: dataset/tokenizer/tokenizer_rhythm.json
  lr: 0.001

# data config
data:
  root_dir: dataset/RealCamera
  img_suffix: .jpg
  label_suffix: _formatted.semantic
  max_len: 64
  batch_size: 1
  num_workers: 4
  pin_memory: true
  shuffle: true
ckpt_path: null