# lightning.pytorch==2.4.0
seed_everything: 42
trainer:
  accelerator: auto
  strategy:
    class_path: lightning.pytorch.strategies.DDPStrategy
    init_args:
      find_unused_parameters: true
  devices: auto
  num_nodes: 1
  precision: 32
  logger:
    class_path: lightning.pytorch.loggers.WandbLogger
    init_args:
      name: isoform_expression
      save_dir: logs
      project: rna_tasks
  callbacks:
  - class_path: lightning.pytorch.callbacks.ModelCheckpoint
    init_args:
      dirpath: logs/
      filename: best_val:{step}-{val_spearman:.3f}-{val_r2:.3f}
      monitor: val_spearman
      save_last: true
      save_top_k: 1
      mode: max
      every_n_train_steps: 500
      every_n_epochs: null
      save_on_train_epoch_end: true
  max_epochs: 3
  max_steps: -1
  val_check_interval: 499
  check_val_every_n_epoch: null
  log_every_n_steps: 1
  accumulate_grad_batches: 1
  gradient_clip_val: 0.1
  gradient_clip_algorithm: null
  default_root_dir: logs
model:
  class_path: modelgenerator.tasks.MMSequenceRegression
  init_args:
    backbone:
      class_path: modelgenerator.backbones.enformer
      init_args:
        max_length: 196_608
        frozen: false
    backbone1:
      class_path: modelgenerator.backbones.aido_rna_1b600m_cds
      init_args:
        max_length: 1024
        frozen: false 
        use_peft: true  
        save_peft_only: true 
        lora_r: 32
        lora_alpha: 64
        lora_dropout: 0.1
        lora_target_modules:
        - query
        - value
        config_overwrites:
          hidden_dropout_prob: 0
          attention_probs_dropout_prob: 0
        model_init_args: null
    backbone_order:
    - dna_seq
    - rna_seq
    adapter:
      class_path: modelgenerator.adapters.fusion.MMFusionTokenAdapter
      init_args:
        fusion:
          class_path: modelgenerator.adapters.fusion.ConcatFusion
          init_args:
            project_size: 1024
            pooling: mean_pooling
        adapter:
          class_path: modelgenerator.adapters.MLPAdapter
          init_args:
            hidden_sizes:
            - 1024
            bias: true
            dropout: 0.1
            dropout_in_middle: false
    num_outputs: 30
    optimizer:
      class_path: torch.optim.AdamW
      init_args:
        lr: 1e-4
        betas:
        - 0.9
        - 0.999
        eps: 1.0e-08
        weight_decay: 0.01
        amsgrad: false
        maximize: false
        foreach: null
        capturable: false
        differentiable: false
        fused: null
    lr_scheduler:
      class_path: modelgenerator.lr_schedulers.CosineWithWarmup
      init_args:
        warmup_ratio: 0.01
        num_warmup_steps: null
        last_epoch: -1
        verbose: deprecated
    use_legacy_adapter: false
    strict_loading: true
    reset_optimizer_states: false
data:
  class_path: modelgenerator.data.IsoformExpression
  init_args:
    path: genbio-ai/transcript_isoform_expression_prediction
    config_name: null
    valid_split_name: valid
    train_split_files:
    - train_*.tsv
    valid_split_files:
    - validation.tsv
    test_split_files:
    - test.tsv
    x_col:
    - rna_seq
    - dna_seq
    normalize: true
    random_seed: 42
    batch_size: 2
    shuffle: true
    sampler: null
    num_workers: 0
    pin_memory: true
    persistent_workers: false
    cv_num_folds: 1
    cv_test_fold_id: 0
    cv_enable_val_fold: true
    cv_fold_id_col: null
ckpt_path: null