Setting parameters of Trainer returned by .get_trainer()

[Tomiinek]

First, thank you for the current Trainer changes in #519!

I cannot figure out how to set trainer parameters when using .get_trainer(). The PR mentions an option:

trainer = tokenizer.model.get_trainer()
trainer.vocab_size = s
trainer.special_tokens = st
tokenizer.train(files, trainer=trainer)

But it does not work for me and says: AttributeError: 'WordLevelTrainer' object has no attribute 'vocab_size'

Is there a way to do this? I can use trainer = type(trainer)(vocab_size=s, special_tokens=st), but it does not seem to be the preferred way.

[n1t0]

Indeed, this is something that we are currently adding, it will be possible as soon as we merge #530

In the meantime, your workaround is necessary.

[Tomiinek]

Ok, thanks!