edu.stanford.nlp.loglinear.storage

Class ModelBatch

java.lang.Object

java.util.AbstractCollection<E>

java.util.AbstractList<E>

java.util.ArrayList<GraphicalModel>

edu.stanford.nlp.loglinear.storage.ModelBatch

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, java.lang.Iterable<GraphicalModel>, java.util.Collection<GraphicalModel>, java.util.List<GraphicalModel>, java.util.RandomAccess

public class ModelBatch
extends java.util.ArrayList<GraphicalModel>

Created on 10/17/15.

Author:

keenon

The idea here is pretty straightforward, but requires some explanation.

GraphicalModels are great for storing lots of metadata about the model, though storing full featurizations can be a bit slow.

With a ModelBatch, you can get your models from anywhere, and after running LENSE on them (which will add lots of annotations, potentially) you can write those models to disk in a big fat batch. Those models you've stored can be stored without featurizing them, as long as you keep enough metadata to be able to featurize later. Then when you load a batch from disk to run simulations, you can try out different feature sets and gameplayers, all while keeping the beautifully precomputed metadata for the model (including instructions for querying, and the query logs).

See Also:

Serialized Form

Field Summary

Fields inherited from class java.util.AbstractList

modCount

Constructor Summary

Constructors

Constructor and Description
ModelBatch() Creates an empty ModelBatch
ModelBatch(java.io.InputStream inputStream) Load a batch of models from disk, without specifying a function to re-featurize those models.
ModelBatch(java.io.InputStream inputStream, java.util.function.Consumer<GraphicalModel> featurizer) Load a batch of models from disk, while running the function "featurizer" on each of the models before adding it to the batch.
ModelBatch(java.lang.String filename) This loads a model batch from a file, then closes the file handler.
ModelBatch(java.lang.String filename, java.util.function.Consumer<GraphicalModel> featurizer) This loads a model batch from a file, then closes the file handler.

Method Summary

Modifier and Type	Method and Description
void	writeToFile(java.lang.String filename) Convenience function to write the current state of the modelBatch out to a file, including all factors.
void	writeToFileWithoutFactors(java.lang.String filename) Convenience function to write the current state of the modelBatch out to a file, without factors.
void	writeToStream(java.io.OutputStream outputStream) This writes the entire batch, including all factors, to the given output stream.
void	writeToStreamWithoutFactors(java.io.OutputStream outputStream) This writes the whole batch, WITHOUT FACTORS, which means that anyone loading this batch will need to include their own featurizer.

Methods inherited from class java.util.ArrayList

add, add, addAll, addAll, clear, clone, contains, ensureCapacity, forEach, get, indexOf, isEmpty, iterator, lastIndexOf, listIterator, listIterator, remove, remove, removeAll, removeIf, removeRange, replaceAll, retainAll, set, size, sort, spliterator, subList, toArray, toArray, trimToSize

Methods inherited from class java.util.AbstractList

equals, hashCode

Methods inherited from class java.util.AbstractCollection

containsAll, toString

Methods inherited from class java.lang.Object

finalize, getClass, notify, notifyAll, wait, wait, wait

Methods inherited from interface java.util.List

containsAll, equals, hashCode

Methods inherited from interface java.util.Collection

parallelStream, stream

Constructor Detail

ModelBatch

public ModelBatch()

Creates an empty ModelBatch

ModelBatch

public ModelBatch(java.lang.String filename)
           throws java.io.IOException

This loads a model batch from a file, then closes the file handler. Just a convenience.

Parameters:

filename - the file to load from

Throws:

java.io.IOException

ModelBatch

public ModelBatch(java.lang.String filename,
                  java.util.function.Consumer<GraphicalModel> featurizer)
           throws java.io.IOException

This loads a model batch from a file, then closes the file handler. Just a convenience.

Parameters:

filename - the file to load from

featurizer - a function that gets run on every GraphicalModel, and has a chance to edit them (eg by adding or changing features)

Throws:

java.io.IOException

ModelBatch

public ModelBatch(java.io.InputStream inputStream)
           throws java.io.IOException

Load a batch of models from disk, without specifying a function to re-featurize those models.

Parameters:

inputStream - the inputstream to load from

Throws:

java.io.IOException

ModelBatch

public ModelBatch(java.io.InputStream inputStream,
                  java.util.function.Consumer<GraphicalModel> featurizer)
           throws java.io.IOException

Load a batch of models from disk, while running the function "featurizer" on each of the models before adding it to the batch. This gives the loader a chance to experiment with new featurization techniques.

Parameters:

inputStream - the input stream to load from

featurizer - a function that gets run on every GraphicalModel, and has a chance to edit them (eg by adding or changing features)

Throws:

java.io.IOException

Method Detail

writeToFile

public void writeToFile(java.lang.String filename)
                 throws java.io.IOException

Convenience function to write the current state of the modelBatch out to a file, including all factors.

WARNING: These files can get quite large, if you're using large embeddings as features.

Parameters:

filename - the file to write the batch to

Throws:

java.io.IOException

writeToFileWithoutFactors

public void writeToFileWithoutFactors(java.lang.String filename)
                               throws java.io.IOException

Convenience function to write the current state of the modelBatch out to a file, without factors.

Parameters:

filename - the file to write the batch to

Throws:

java.io.IOException

writeToStream

public void writeToStream(java.io.OutputStream outputStream)
                   throws java.io.IOException

This writes the entire batch, including all factors, to the given output stream.

WARNING: These files can get quite large, if you're using large embeddings as features.

Parameters:

outputStream - the outputstream to write our files to

Throws:

java.io.IOException

writeToStreamWithoutFactors

public void writeToStreamWithoutFactors(java.io.OutputStream outputStream)
                                 throws java.io.IOException

This writes the whole batch, WITHOUT FACTORS, which means that anyone loading this batch will need to include their own featurizer. Make sure that you have sufficient metadata to be able to do full featurizations.

Parameters:

outputStream - the outputstream to write our files to

Throws:

java.io.IOException