def create_inference_dataset(self, dataset, tag, config,
training_set_metadata):
if self.backend.df_engine.partitioned:
return PartitionedDataset(
dataset, get_proc_features(config),
training_set_metadata.get(DATA_TRAIN_HDF5_FP))
else:
return PandasDataset(dataset, get_proc_features(config),
training_set_metadata.get(DATA_TRAIN_HDF5_FP))
def create_inference_dataset(self, dataset, tag, config,
training_set_metadata):
"""We don't use TFRecord for inference."""
if self.backend.df_engine.partitioned:
raise ValueError(
'Batch inference not supported with TFRecord format at this time'
)
else:
return PandasDataset(dataset, get_proc_features(config),
training_set_metadata.get(DATA_TRAIN_HDF5_FP))
def create_dataset(self, dataset, tag, config, training_set_metadata):
return Dataset(dataset, get_proc_features(config),
training_set_metadata.get(DATA_TRAIN_HDF5_FP))
def create(self, dataset: Union[str, DataFrame], config: Dict[str, Any],
training_set_metadata: Dict[str, Any]):
return RayDataset(dataset, get_proc_features(config),
training_set_metadata, self.backend)
def create(self, dataset: DataFrame, config: Dict[str, Any],
training_set_metadata: Dict[str, Any]):
return RayDataset(dataset, get_proc_features(config),
training_set_metadata)