Skip to main content
Version: Experimental 🚧

Train model

The instance created by the Main class allow to train a model, in this section we will see how to use the train method.

The train method is used to train a model. Has the following parameters:

  • model_data (dictionary): Data of the model to train, this data is returned by the add_model method.
  • dataset_name (string): Name of the dataset to train the model.
  • epochs (int, optional): Number of epochs to train the model. By default will be used the value defined in the epochs parameter of the add_dataset method.
  • batch_size (int, optional): Batch size to train the model. By default will be used the value defined in the batch_size parameter of the add_dataset method.
  • initial_epoch (int, optional): Initial epoch to train the model. By default is 0.
  • shuffle_buffer (int, optional): Size of the shuffle buffer. By default will be used the value defined in the shuffle_buffer parameter of the add_dataset method.
  • force_creation (bool, optional): If is True the model will be create again even if the model already exists. By default is False.
  • train_ds (tf.data.Dataset | List, optional): Dataset to train the model. By default will be used the dataset defined in the add_dataset method.
  • val_ds (tf.data.Dataset | List, optional): Dataset to validate the model. By default will be used the dataset defined in the add_dataset method.
Example
ia_maker.train(
  model_1_data,
  epochs=5,
  dataset_name='dataset_1'
)
info

If a model is already trained and you call the train method again, the model will be loaded and continue to train, but don't forget to send the initial_epoch parameter with the number of epochs already trained.