TypeError: Expected maxsize to be an integer or None

[Adibian]

I tried to train FastPitch in google colab and I followed the Quick Start Guide and after downloading the LJSpeech-1.1 dataset when executing the following command:
bash scripts/prepare_dataset.sh

I get the this error:

Processing filelists/ljs_audio_text.txt...
Traceback (most recent call last):
  File "prepare_dataset.py", line 174, in <module>
    main()
  File "prepare_dataset.py", line 129, in main
    pitch_online_method=args.f0_method)
  File "/content/DeepLearningExamples/PyTorch/SpeechSynthesis/FastPitch/fastpitch/data_function.py", line 190, in __init__
    self.betabinomial_interpolator = BetaBinomialInterpolator()
  File "/content/DeepLearningExamples/PyTorch/SpeechSynthesis/FastPitch/fastpitch/data_function.py", line 54, in __init__
    self.bank = functools.lru_cache(beta_binomial_prior_distribution)
  File "/usr/lib/python3.7/functools.py", line 490, in lru_cache
    raise TypeError('Expected maxsize to be an integer or None')
TypeError: Expected maxsize to be an integer or None

[alancucki]

Hi @majidAdibian77 ,

thanks for reporting this bug!
You're using Python <= 3.7, in Python 3.8+ lru_cache has gotten user_function argument, which is used in FastPitch code.

In the Quick Start Guide, we recommend the setup that we support using NGC containers, which come with up-to-date Python.

[Adibian]

Hi @alancucki
Thanks for your answer.
Yes, you are right.
Also I could solve this problem without updating python.
I replaced following line in fastpitch/data_function.py/BetaBinomialInterpolator class/__init__ function (this line):
self.bank = functools.lru_cache(beta_binomial_prior_distribution)

with two new lines:

f = functools.lru_cache(maxsize=128)
self.bank = f(beta_binomial_prior_distribution)

and then this error was gone :)

[DarkIndoraptor]

One question, does this also work with lutris?