# @package _group_
type: LitVQVAEModel
args:
    optimizer:
        type: MADGRAD
        args:
            lr: 1.0e-3
            momentum: 0.9
            weight_decay: 0
            eps: 1.0e-6
    lr_scheduler:
        type: OneCycleLR
        args:
            interval: step
            max_lr: 1.0e-3
            three_phase: true
            epochs: 64
            steps_per_epoch: 633 # num_samples / batch_size
    criterion:
        type: MSELoss
        args:
            reduction: mean
    monitor: val_loss
    mapping: sentence_piece