Package ai.djl.training.dataset
Class DataIterable
java.lang.Object
ai.djl.training.dataset.DataIterable
- Direct Known Subclasses:
BulkDataIterable
DataIterable is a data loader that combines
Dataset
, Batchifier
, Pipeline
, and Sampler
to provide an iterable over the given RandomAccessDataset
.
We don't recommended using DataIterable directly. Instead use RandomAccessDataset
combined with Trainer
to iterate over the RandomAccessDataset
}
-
Field Summary
FieldsModifier and TypeFieldDescriptionprotected Batchifier
protected RandomAccessDataset
protected Device
protected Batchifier
protected NDManager
protected Pipeline
protected Pipeline
-
Constructor Summary
ConstructorsConstructorDescriptionDataIterable
(RandomAccessDataset dataset, NDManager manager, Sampler sampler, Batchifier dataBatchifier, Batchifier labelBatchifier, Pipeline pipeline, Pipeline targetPipeline, ExecutorService executor, int preFetchNumber, Device device) Creates a new instance ofDataIterable
with the given parameters. -
Method Summary
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
Methods inherited from interface java.lang.Iterable
forEach, spliterator
Methods inherited from interface java.util.Iterator
forEachRemaining, remove
-
Field Details
-
dataset
-
manager
-
dataBatchifier
-
labelBatchifier
-
pipeline
-
targetPipeline
-
device
-
-
Constructor Details
-
DataIterable
public DataIterable(RandomAccessDataset dataset, NDManager manager, Sampler sampler, Batchifier dataBatchifier, Batchifier labelBatchifier, Pipeline pipeline, Pipeline targetPipeline, ExecutorService executor, int preFetchNumber, Device device) Creates a new instance ofDataIterable
with the given parameters.- Parameters:
dataset
- the dataset to iterate onmanager
- the manager to create the arrayssampler
- a sampler to sample data withdataBatchifier
- a batchifier for datalabelBatchifier
- a batchifier for labelspipeline
- the pipeline of transforms to apply on the datatargetPipeline
- the pipeline of transforms to apply on the labelsexecutor
- anExecutorService
preFetchNumber
- the number of samples to prefetchdevice
- theDevice
-
-
Method Details