def __init__(self, args):
""" Initializes the task.
Args:
args: A dict of model configurations.
"""
src_data_pipeline_cls = args.get("src_data_pipeline.class", TextDataPipeline)
src_data_pipeline_params = args.get("src_data_pipeline.params", None) or {}
self._src_data_pipeline = build_data_pipeline(
src_data_pipeline_cls, **src_data_pipeline_params)
trg_data_pipeline_cls = args.get("trg_data_pipeline.class", TextDataPipeline)
trg_data_pipeline_params = args.get("trg_data_pipeline.params", None) or {}
self._trg_data_pipeline = build_data_pipeline(
trg_data_pipeline_cls, **trg_data_pipeline_params)
super(Seq2Seq, self).__init__(args)
def __init__(self, args):
""" Initializes with configuration. """
super(MultiTaskSpeechTranslation, self).__init__(args)
transcript_dp_cls = args.get("transcript_data_pipeline.class",
TranscriptDataPipeline)
transcript_dp_params = args.get("transcript_data_pipeline.params",
None) or {}
self._transcript_data_pipeline = build_data_pipeline(
transcript_dp_cls, **transcript_dp_params)
translation_dp_cls = args.get("translation_data_pipeline.class",
TranscriptDataPipeline)
translation_dp_params = args.get("translation_data_pipeline.params",
None) or {}
self._translation_data_pipeline = build_data_pipeline(
translation_dp_cls, **translation_dp_params)
def __init__(self, args):
""" Initializes the task.
Args:
args: A dict of model configurations.
"""
data_pipeline_cls = args.get("data_pipeline.class", TextDataPipeline)
data_pipeline_params = args.get("data_pipeline.params", None) or {}
self._data_pipeline = build_data_pipeline(data_pipeline_cls,
**data_pipeline_params)
self._begin_of_sentence = args.get("begin_of_sentence", "bos")
super(LanguageModel, self).__init__(args)
def __init__(self, args):
""" Initializes the task.
Args:
args: A dict of model configurations.
"""
super(SpeechToText, self).__init__(args)
trg_data_pipeline_cls = args.get("transcript_data_pipeline.class",
TranscriptDataPipeline)
trg_data_pipeline_params = args.get("transcript_data_pipeline.params",
None) or {}
self._trg_data_pipeline = build_data_pipeline(
trg_data_pipeline_cls, **trg_data_pipeline_params)
self._audio_feature_dim = args["audio_feature_dim"]
self._audio_feature_channels = args["audio_feature_channels"]
self._specaug = SpecAugment.build(args.get("specaug", None))
