espnet2.speechlm.dataloader.multimodal_loader.text_loader.OmniIOTextReader
Less than 1 minute
espnet2.speechlm.dataloader.multimodal_loader.text_loader.OmniIOTextReader
class espnet2.speechlm.dataloader.multimodal_loader.text_loader.OmniIOTextReader(parquet_path: str, valid_ids: list | None = None)
Bases: object
Dict-like lazy text reader using omniio parquets.
Reads compressed text data from omniio parquet files. Text is stored in binary format with compression and accessed via byte offsets.
- Parameters:
- parquet_path – Path to the parquet file containing text metadata.
- valid_ids – List of valid IDs to keep (optional, keeps all if None).
items() → Iterator[Tuple[str, str]]
Return iterator over (id, text) pairs.
keys() → Iterator[str]
Return iterator over IDs.
values() → Iterator[str]
Return iterator over text values.
