| udpos --learning_rate=4.5e-5 --eval_steps=1000 --per_device_batch_size=10 --max_steps=1000 language_source=top8lang1 language_target=tl |
| udpos --learning_rate=4.5e-5 --eval_steps=1000 --per_device_batch_size=10 --max_steps=1000 language_source=top8lang1 language_target=tl |