espnet3.systems.f5tts.src.pinyin.load_vocab_char_map

source

espnet3.systems.f5tts.src.pinyin.load_vocab_char_map(vocab_file: str) → Dict[str, int]

Load F5’s vocab.txt as {token: index} (line number = index).

• Parameters:vocab_file – Path to F5’s vocab.txt, one token per line.
• Returns: Mapping from token to its line number.

Example

>>> load_vocab_char_map("vocab.txt")   # first line is a space
{' ': 0, 'a': 1, 'b': 2}

NOTE

Only the trailing newline is stripped, never surrounding whitespace, so F5’s literal space token at index 0 survives. Stripping it would shift every id and silently corrupt the mapping. Unlike an ESPnet token list this vocabulary has no <blank> / <unk> / <sos/eos>.