char-level Tokenizer.sequences_to_texts() insert additional SPACEs

[XiYuan68]

• Check that you are up-to-date with the master branch of keras-preprocessing. You can update with:
  pip install git+git://github.com/keras-team/keras-preprocessing.git --upgrade --no-deps

• Provide a link to a GitHub Gist of a Python script that can reproduce your issue (or just copy the script here if it is short).

Describe the problem.

from tensorflow.keras.preprocessing.text import Tokenizer

text = ['abc def']
tokenizer = Tokenizer(char_level=True, split='')
tokenizer.fit_on_texts(text)
sequence = tokenizer.texts_to_sequences(text)
text_after = tokenizer.sequences_to_texts(sequence)

print(text_after)
>>> ['a b c   d e f']

notice that text_after and text are different, additional SPACEs are inserted

Describe the expected behavior.

text_after should be same as text

I believe this line is where the problem is, replacing:

vect = ' '.join(vect)

with

vect = self.split.join(vect)

will fix the bug in my mini case