espnet2.speechlm.dataloader.multimodal_loader.dialogue_loader.OmniIODialogueLoader
Less than 1 minute
espnet2.speechlm.dataloader.multimodal_loader.dialogue_loader.OmniIODialogueLoader
class espnet2.speechlm.dataloader.multimodal_loader.dialogue_loader.OmniIODialogueLoader(parquet_path: str, valid_ids: list | None = None)
Bases: OmniIOTextReader
Dict-like lazy dialogue reader using omniio parquets.
Extends OmniIOTextReader to parse JSON string output into validated dialogue messages with the same sanity checks as DialogueReader.
- Parameters:
- parquet_path – Path to the parquet file containing dialogue metadata.
- valid_ids – List of valid IDs to keep (optional, keeps all if None).
- worker_id – Partition IDs by worker (optional, keeps all if None).
- world_size – Used for worker partitioning.
items() → Iterator[Tuple[str, List[Tuple[str, str, str | Tuple[ndarray, int]]]]]
Return iterator over (id, validated_dialogue) pairs.
values() → Iterator[List[Tuple[str, str, str | Tuple[ndarray, int]]]]
Return iterator over validated dialogue values.
