model_module: transformer data: unit_type: spm model: n_vocab: 5002 pos_enc: null embed_unit: 128 att_unit: 512 head: 8 unit: 2048 layer: 16 dropout_rate: 0.5 emb_dropout_rate: 0.0 att_dropout_rate: 0.0 tie_weights: False decoding: batch_size: 30 num_workers: 2