template.yaml

# @package _global_
# A config template for running Weasel on your own data and end-model
#   ??? means that the field must be defined.

seed: 7
ignore_warnings: True

datamodule:
  _target_: ???
  # The following three params are ignored by the data_module, but semantically are tied to the dataset,
  # which is why they are included here. In practice though, only Weasel needs to know them, see below in the Weasel params
  # where they are all automatically inferred from these params in datamodule.
  num_LFs: ???
  n_classes: ???
  class_balance: null # will assume uniform classes if null
  # Actual params used by DataModule
  batch_size: 64
  val_test_split: [250, -1]
  num_workers: 0
  pin_memory: True
  seed: 3

trainer:
  _target_: pytorch_lightning.Trainer
  gpus: 0  # set to 0 to train on CPU, and >= 1 for GPU(s)
  min_epochs: 1
  max_epochs: 100
  # resume_from_checkpoint: <ckpt_dir>/best.ckpt

end_model:
  _target_: ???
  # What comes here really depends on the end-model!
  # dropout: 0.3
  # net_norm: "none"
  # activation_func: "GeLU"
  # input_dim: 300
  # hidden_dims: [50, 50, 25]
  output_dim: ${datamodule.n_classes}
  # adjust_thresh: True


Weasel:
  _target_: weasel.models.weasel.Weasel
   # very convenient interpolation by Hydra/OmegaConf that infers params from the dataset details
  num_LFs: ${datamodule.num_LFs}
  n_classes: ${datamodule.n_classes}
  class_balance: ${datamodule.class_balance}
  monitor: "Val/auc"  # adjust for multi-class case

  loss_function: "cross_entropy"

  temperature: 2.0
  accuracy_scaler: "sqrt"

  use_aux_input_for_encoder: False
  class_conditional_accuracies: True

  encoder:
    _target_: weasel.models.encoder_models.encoder_MLP.MLPEncoder
    # example_extra_input: [1, 300] # OR (1, ${end_model.input_dim})
    #  |--> Commented out, since it is better to specify the 'example_input_array' attribute in your end-model
    dropout: 0.3
    net_norm: "batch_norm"
    activation_func: "GeLU"
    hidden_dims: [70, 70]  # although you probably want to different hidden dim. configurations!

  optim_end_model:
    _target_: torch.optim.Adam
    lr: 1e-4
    weight_decay: 7e-7

  optim_encoder:
    _target_: torch.optim.Adam
    lr: 1e-4
    weight_decay: 7e-7

  scheduler:
    _target_: torch.optim.lr_scheduler.ReduceLROnPlateau
    mode: "max"
    factor: 0.5
    patience: 10

  verbose: True

callbacks:
  model_checkpoint:
    _target_: pytorch_lightning.callbacks.ModelCheckpoint
    monitor: ${Weasel.monitor}
    save_top_k: 2
    save_last: True
    mode: "max"
    dirpath: "checkpoints/"
    filename: "{epoch:02d}"

logger:
  wandb:
    _target_: pytorch_lightning.loggers.wandb.WandbLogger
    # entity: ""  # optionally set to name of your wandb team
    name: null
    tags: []
    notes: "..."
    project: "Weasel"   # give it a better name ;)
    group: ""
    mode: online # disabled  # disabled for no wandb logging