TOKENIZERS_PARALLELISM

[fstairu]

I'm trying to run FluxGym on Runpod and constantly experience the issues. The last, and guess the least one (!) is
Avoid using tokenizers before the fork if possible
[2025-01-26 12:10:31] [INFO] - Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
And that's the training has not yet been started, just the pre-sample was generated ("save_at_first" enabled on advanced setup)
Absolutely have no clue how to fix it, and have no any desire to spent the money for trying this thing fly, just praying it will not resemble into any other issues!

[Racaya]

You can either wait a few minutes and the script should continue (at it always did for me).

Or, you can add the environment variable at the top of app.py, where the other env variables are defined:
os.environ['TOKENIZERS_PARALLELISM'] = 'False'

This resolved the problem for me.