`nemo_rl.evals.answer_parsing`#

Contains utility functions for answer parsing.

Module Contents#

Functions#

`normalize_extracted_answer`
`normalize_response`	Normalize the response by removing markdown and LaTeX formatting that may prevent a match.

Data#

`MULTILINGUAL_ANSWER_PATTERN_TEMPLATE`
`MULTILINGUAL_ANSWER_REGEXES`

API#

nemo_rl.evals.answer_parsing.MULTILINGUAL_ANSWER_PATTERN_TEMPLATE#: ‘(?i){}[ \t]*([A-D]|[أ-د]|[অ]|[ব]|[ড]|[ঢ]|[Ａ]|[Ｂ]|[Ｃ]|[Ｄ])’

nemo_rl.evals.answer_parsing.MULTILINGUAL_ANSWER_REGEXES#: [‘Answer\s*:’, ‘Answer\s*:\u200b\u200b\u200b\u200b\u200b\u200b’, ‘উত্তর\s*:’, ‘उत्तर\s*:’, ‘উত্ত…

nemo_rl.evals.answer_parsing.normalize_extracted_answer(extracted_answer: str) → str[source]#

nemo_rl.evals.answer_parsing.normalize_response(response: str) → str[source]#: Normalize the response by removing markdown and LaTeX formatting that may prevent a match.