Class MLIntegrationExamples.TensorFlowAdapter

java.lang.Object

neqsim.process.safety.risk.ml.MLIntegrationExamples.BaseMLAdapter

neqsim.process.safety.risk.ml.MLIntegrationExamples.TensorFlowAdapter

All Implemented Interfaces:

Serializable, MLIntegrationExamples.MLModelAdapter

Enclosing class:

MLIntegrationExamples

public static class MLIntegrationExamples.TensorFlowAdapter
extends MLIntegrationExamples.BaseMLAdapter

Adapter for TensorFlow SavedModel format.

Loads TensorFlow models using the TensorFlow Java API.

Dependencies Required

<dependency>
  <groupId>org.tensorflow</groupId>
  <artifactId>tensorflow-core-platform</artifactId>
  <version>0.5.0</version>
</dependency>

Since:

3.3.0

Version:

1.0

Author:

NeqSim Development Team

See Also:

• Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
private String	inputTensorName
private String	modelDir
private String	outputTensorName
private static final long	serialVersionUID

Fields inherited from class MLIntegrationExamples.BaseMLAdapter

featureDefaults, inputFeatures, isLoaded, modelName

Constructor Summary

Constructors

Constructor	Description
TensorFlowAdapter(String modelDir, List<String> inputFeatures, String inputTensorName, String outputTensorName)	Creates a TensorFlow adapter.

Method Summary

Modifier and Type	Method	Description
void	load()	Loads the TensorFlow model.
double	predict(Map<String,Double> features)	Predicts output from input features.

Methods inherited from class MLIntegrationExamples.BaseMLAdapter

getInputFeatures, getModelName, isLoaded, prepareInput

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

serialVersionUID

private static final long serialVersionUID

See Also:

• Constant Field Values

modelDir

private String modelDir

inputTensorName

private String inputTensorName

outputTensorName

private String outputTensorName

Constructor Details

TensorFlowAdapter

public TensorFlowAdapter(String modelDir,
 List<String> inputFeatures,
 String inputTensorName,
 String outputTensorName)

Creates a TensorFlow adapter.

Parameters:

modelDir - path to SavedModel directory

inputFeatures - input feature names

inputTensorName - name of input tensor

outputTensorName - name of output tensor

Method Details

load

public void load()

Loads the TensorFlow model.

In production:

model = SavedModelBundle.load(modelDir, "serve");

predict

public double predict(Map<String,Double> features)

Description copied from interface: MLIntegrationExamples.MLModelAdapter

Predicts output from input features.

Parameters:

features - input features as name-value map

Returns:

prediction score or probability