Guoao Wei
Guoao Wei
Just add custom Series type to align with the standard `preprocessing.py`. I am thinking simply not to call functions or variables from other modules, unless that module is language agnostic....
> For the rest: I agree! Should I create another PR? That issue should be solved before this PR could continue.
@henrifroese Do you think we should proceed with that?
> It's difficult for us to not to import functions from other modules at the moment (e.g. I'm not sure how we would not use the tokenize function in representation...
Why does the error `The command "black --check ." exited with 1.` occurs in CI? I ran this check locally and there's nothing wrong.
Just fix the problem, thank you! @jbesomi
Agreed. I want to use a pre-trained mode for my further research. Training one seems too time-consuming.
@JonathanRaiman That would be great! Looking forward to a pre-trained model.
I'm also facing the same question. Any updates on this issue?
@ThomasRetornaz @david-macleod I just found a workaround if you're interested. You can take out each element in this `Tuple[List[Tensor]]`, feed them to the model separately, and put them back into...