speakleash / speakleash-instruct-creator

Generate instructions datasets for the fine-tuning purposes.
3 stars 5 forks source link

Fix create_password.py - "litery" instead of "znaki" #40

Closed jansowa closed 4 months ago

jansowa commented 5 months ago

In the "create_password.py" script, the word "characters" ("znaki") is used in the instructions, but the responses show that "letters" ("litery") are counted.

The word "characters" should be replaced with "letters".

By the way, you can try to create more diverse instructions.

jansowa commented 5 months ago

I'll do that one :)

jansowa commented 5 months ago

There are even more problems...

  1. Generating invalid passwords. Sample:
    {
    "instruct": "Wygeneruj losowe hasło składające się z 14 liter, 2 cyfr i 5 znaków specjalnych.",
    "input": "",
    "output": "'y\"!!'r{03[}3W55|&+((",
    "source_name": "password_creator",
    "source_url": "https://speakleash.org/",
    "source_description": "Instrukcje zostały wygenerowane automatycznie",
    "script_name": "create_password.py"
    }
  2. Generating invalid instructions, for example: "Stwórz hasło łatwe do zapamiętania składające się 14 liter, 2 cyfr i 5 znaków specjalnych." (without "z"). There are also some other similar problems.

I'll fix all of them.

jansowa commented 5 months ago

@chrisociepa suggested that ultimately these instructions might not be useful anyway, so I'm just uploading what I've already done - without fix of the bug associated with generating invalid passwords.

IgorTest19 commented 4 months ago

Changes were merged. Closing issue.