Model section

In the model section, you should define the model to use (and its parameters), along with the training parameters. Note: The model to use must be the first attribute of the model section!

Model section

model:
  MODEL_TO_USE:
    PARAM1: VAL1
    PARAM2:
      - VAL2
      - VAL3
    ...


  # Sequence of tasks to use during the training phase of the model.
  # Each sample of the train dataset will be applied to one of the followings or
  # all of them, depending on the `train_task_selection_strat` parameter
  #
  # Required
  train_tasks:
    - SequentialSideInfoTask
    - RatingPredictionTask

  # When training according to the multitask paradigm,
  # there are two different strategies available for choosing the task to apply
  # for the particular sample of the training set currently processed:
  # - "all" will apply, for the particular sample, ALL training tasks defined # (2)
  # - "random" will apply, for the particular sample, ONE training task among those defined choosen at random # (3)
  #
  # Optional, Default: "all"
  train_task_selection_strat: all

  # If set, the validation phase is performed at the end of each epoch of training:
  # - In this case, the best model will be saved according to the `monitor_metric` value
  # val_task should be set to one of the available tasks
  #
  # Optional, Default: null
  val_task: null

  # If `val_task` parameter is set, and thus the validation phase is performed,
  # you could specify the exact template of the `val_task` to use for validation.
  # If `val_task` is set but this parameter is set to null, then validation is performed
  # by choosing random templates of the `val_task` to apply to each sample of the
  # val dataset
  #
  # Optional, Default: null
  val_task_template_id: null

  # Number of epochs to perform during the training phase
  #
  # Optional, Default: 10
  n_epochs: 10

  # If `val_task` is set, and thus the validation phase is performed,
  # you can change the metric that should be used in order to save the best model
  #
  # Optional, Default: loss
  monitor_metric: loss

  # The batch size to use during the training phase
  #
  # Optional, Default: 4
  train_batch_size: 4

  # If `val_task` parameter is set, and thus the validation phase is performed,
  # you could change the batch size to use for the validation phase.
  # If this parameter is set to null, then the `train_batch_size` value will be used
  # as the `eval_batch_size`
  eval_batch_size: null

All parameters of the model section should be defined as attribute of the model mapping

Check the available models to see which models are implemented at the moment and their customizable parameters!