espnet2.train.dataset.ESPnetMultiTaskDataset

source

class espnet2.train.dataset.ESPnetMultiTaskDataset(path_name_type_list: Collection[Tuple[str, str, str]], key_file: str | None = None, **kwargs)

Bases: AbsDataset

The top-level Dataset object that can manage multiple EspnetSpeechLMDataset objects, each of which serves a specific task and dataset. This object will query all these EspnetSpeechLMDataset and combine examples from different tasks for multi-task training. Typically, this dataset is used in ESPnet SpeechLM models See details in: <espnet>/egs2/TEMPLATE/speechlm1#data-loading-and-preprocessing

has_name(name) → bool

names() → Tuple[str, ...]