fluid_1.2 tokenizer.py --test_data_dir ./test.txt.utf8 --batch_size 1 > res