Skip to content

configure transcribe() to use langID #15853

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
ealbasiri wants to merge 1 commit into
base: main
Choose a base branch
from
+6 −2
Draft

configure transcribe() to use langID #15853

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion nemo/collections/asr/data/audio_to_text_lhotse_prompt_index.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,6 +83,9 @@ def __init__(self, tokenizer: TokenizerSpec, cfg: Dict) -> None:
# Index used for the language-agnostic / auto prompt
self.auto_index = self.prompt_dict.get('auto', 101)

# Fallback language for cuts with no language (e.g. transcribe() temp cuts)
self.default_lang = cfg.get('default_lang', None)

logging.info(
f"LhotseSpeechToTextBpeDatasetWithPromptIndex: "
f"default_prompt_mode={self.default_prompt_mode}, "
Expand Down Expand Up @@ -121,7 +124,7 @@ def _get_prompt_index_for_cut(self, cut) -> int:
mode = self._get_prompt_mode(cut)

if mode == 'langID':
return self._get_prompt_index(cut.supervisions[0].language)
return self._get_prompt_index(cut.supervisions[0].language or self.default_lang)
elif mode == 'auto':
return self.auto_index
elif mode == 'unified':
Expand Down
3 changes: 2 additions & 1 deletion nemo/collections/asr/models/rnnt_bpe_models_prompt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -235,7 +235,7 @@ def _setup_transcribe_dataloader(self, config: Dict) -> 'torch.utils.data.DataLo
manifest_filepath = os.path.join(config['temp_dir'], 'manifest.json')
batch_size = min(config['batch_size'], len(config['paths2audio_files']))

target_lang = config.get('target_lang', 'en-US')
target_lang = config.get('target_lang', 'auto')

dl_config = {
'manifest_filepath': manifest_filepath,
Expand All @@ -254,6 +254,7 @@ def _setup_transcribe_dataloader(self, config: Dict) -> 'torch.utils.data.DataLo
'num_prompts': self.cfg.model_defaults.get('num_prompts', 128),
'subsampling_factor': self.cfg.get('subsampling_factor', 8),
'default_lang': target_lang,
'default_prompt_mode': 'langID',
'window_stride': self.cfg.preprocessor.get('window_stride', 0.01),
}

Expand Down
Loading