Typo for tokenizer_path arg

[carlosjedwab]

It seems like run_training_4m.py uses the arg text_tokenizer_path to define the path of the text tokenizer, however the config files have this same variable called tokenizer_path. I believe they were supposed to be the same