train.args

udpos -tt=token-classification -tn=udpos28 -mi=xlm-roberta-base -mt=ft --learning_rate=5e-5 --eval_steps=1000 --eval_batch_size=10 --train_batch_size=10 --multi --max_steps=1000