Skip to content

v25.2.0

Choose a tag to compare

@Geekdude Geekdude released this 05 Aug 15:13
· 130 commits to main since this release

25.2.0 - 2025-02-21

Added

  • Added max_training_tokens to specify the max_length parameter for the tokenizer when tokenizing the training set. Defaults to the prior setting of 512.

Changed

  • Logging: Switch from using logging module to loguru.

Fixed

  • load_json_arg(): Fixed bug with json string as input.
  • Fixed bug in interactive run mode.
  • Fixed extract_answer() utility to use the prompt and end string to extract the answer portion.