Package ai.onnxruntime
Class OrtEnvironment
java.lang.Object
ai.onnxruntime.OrtEnvironment
- All Implemented Interfaces:
AutoCloseable
The host object for the ONNX Runtime system. Can create
OrtSessions which encapsulate
specific models. This object should be instantiated before any other ONNX Runtime classes are
created.
There can be at most one OrtEnvironment object created in a JVM lifetime. This class
implements AutoCloseable as before for backwards compatibility with 1.10 and earlier, but
the close method is a no-op. The environment is closed by a JVM shutdown hook registered on
construction.
-
Nested Class Summary
Nested ClassesModifier and TypeClassDescriptionstatic final classControls the global thread pools in the environment. -
Field Summary
FieldsModifier and TypeFieldDescriptionstatic final StringThe default name for ORT environments constructed from Java. -
Method Summary
Modifier and TypeMethodDescriptionvoidclose()Close is a no-op on OrtEnvironment since ORT 1.11.createSession(byte[] modelArray) Create a session using the defaultOrtSession.SessionOptions, model and the default memory allocator.createSession(byte[] modelArray, OrtSession.SessionOptions options) Create a session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.createSession(String modelPath) Create a session using the defaultOrtSession.SessionOptions, model and the default memory allocator.createSession(String modelPath, OrtSession.SessionOptions options) Create a session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.createSession(ByteBuffer modelBuffer) Create a session using the defaultOrtSession.SessionOptions, model and the default memory allocator.createSession(ByteBuffer modelBuffer, OrtSession.SessionOptions options) Create a session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.createTrainingSession(String checkpointPath, String trainPath, String evalPath, String optimizerPath) Create a training session using the defaultOrtSession.SessionOptions, model and the default memory allocator.createTrainingSession(String checkpointPath, String trainPath, String evalPath, String optimizerPath, OrtSession.SessionOptions options) Create a training session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.static EnumSet<OrtProvider>Gets the providers available in this environment.static OrtEnvironmentGets the OrtEnvironment.static OrtEnvironmentgetEnvironment(OrtLoggingLevel logLevel) Gets the OrtEnvironment.static OrtEnvironmentgetEnvironment(OrtLoggingLevel loggingLevel, String name) Gets the OrtEnvironment.static OrtEnvironmentgetEnvironment(OrtLoggingLevel loggingLevel, String name, OrtEnvironment.ThreadingOptions threadOptions) Creates an OrtEnvironment using the specified global thread pool options.static OrtEnvironmentgetEnvironment(String name) Gets the OrtEnvironment.Gets the native library version string.booleanIs training enabled in this build of ONNX Runtime?voidsetTelemetry(boolean sendTelemetry) Turns on or off the telemetry.toString()
-
Field Details
-
DEFAULT_NAME
The default name for ORT environments constructed from Java.- See Also:
-
-
Method Details
-
getEnvironment
Gets the OrtEnvironment. If there is not an environment currently created, it creates one usingDEFAULT_NAMEandOrtLoggingLevel.ORT_LOGGING_LEVEL_WARNING.- Returns:
- The OrtEnvironment singleton.
-
getEnvironment
Gets the OrtEnvironment. If there is not an environment currently created, it creates one using the supplied name andOrtLoggingLevel.ORT_LOGGING_LEVEL_WARNING.If the environment already exists then it returns the existing one and logs a warning if the name or log level is different from the requested one.
- Parameters:
name- The logging id of the environment.- Returns:
- The OrtEnvironment singleton.
-
getEnvironment
Gets the OrtEnvironment. If there is not an environment currently created, it creates one using theDEFAULT_NAMEand the supplied logging level.If the environment already exists then it returns the existing one and logs a warning if the name or log level is different from the requested one.
- Parameters:
logLevel- The logging level to use.- Returns:
- The OrtEnvironment singleton.
-
getEnvironment
Gets the OrtEnvironment. If there is not an environment currently created, it creates one using the supplied name and logging level. If an environment already exists with a different name, that environment is returned and a warning is logged.- Parameters:
loggingLevel- The logging level to use.name- The log id.- Returns:
- The OrtEnvironment singleton.
-
getEnvironment
public static OrtEnvironment getEnvironment(OrtLoggingLevel loggingLevel, String name, OrtEnvironment.ThreadingOptions threadOptions) Creates an OrtEnvironment using the specified global thread pool options. Note unlike the othergetEnvironmentmethods if there already is an existing OrtEnvironment this call throwsIllegalStateExceptionas we cannot guarantee that the environment has the appropriate thread pool configuration.- Parameters:
loggingLevel- The logging level to use.name- The log id.threadOptions- The global thread pool options.- Returns:
- The OrtEnvironment singleton.
-
createSession
Create a session using the defaultOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
modelPath- Path on disk to load the model from.- Returns:
- An
OrtSessionwith the specified model. - Throws:
OrtException- If the model failed to load, wasn't compatible or caused an error.
-
createSession
public OrtSession createSession(String modelPath, OrtSession.SessionOptions options) throws OrtException Create a session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
modelPath- Path on disk to load the model from.options- The session options.- Returns:
- An
OrtSessionwith the specified model. - Throws:
OrtException- If the model failed to load, wasn't compatible or caused an error.
-
createSession
public OrtSession createSession(ByteBuffer modelBuffer, OrtSession.SessionOptions options) throws OrtException Create a session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
modelBuffer- Byte buffer representing an ONNX model. Must be a direct byte buffer.options- The session options.- Returns:
- An
OrtSessionwith the specified model. - Throws:
OrtException- If the model failed to parse, wasn't compatible or caused an error.
-
createSession
Create a session using the defaultOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
modelBuffer- Byte buffer representing an ONNX model. Must be a direct byte buffer.- Returns:
- An
OrtSessionwith the specified model. - Throws:
OrtException- If the model failed to parse, wasn't compatible or caused an error.
-
createSession
public OrtSession createSession(byte[] modelArray, OrtSession.SessionOptions options) throws OrtException Create a session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
modelArray- Byte array representing an ONNX model.options- The session options.- Returns:
- An
OrtSessionwith the specified model. - Throws:
OrtException- If the model failed to parse, wasn't compatible or caused an error.
-
createSession
Create a session using the defaultOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
modelArray- Byte array representing an ONNX model.- Returns:
- An
OrtSessionwith the specified model. - Throws:
OrtException- If the model failed to parse, wasn't compatible or caused an error.
-
createTrainingSession
public OrtTrainingSession createTrainingSession(String checkpointPath, String trainPath, String evalPath, String optimizerPath) throws OrtException Create a training session using the defaultOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
checkpointPath- Path to the checkpoint folder.trainPath- Path to the training model.evalPath- Path to the evaluation model. Null signifies there is no eval model.optimizerPath- Path to the optimizer model. Null signifies there is no optimizer model.- Returns:
- An
OrtTrainingSessionwith the specified model loaded. - Throws:
OrtException- If the model failed to load, wasn't compatible or caused an error.
-
createTrainingSession
public OrtTrainingSession createTrainingSession(String checkpointPath, String trainPath, String evalPath, String optimizerPath, OrtSession.SessionOptions options) throws OrtException Create a training session using the specifiedOrtSession.SessionOptions, model and the default memory allocator.- Parameters:
checkpointPath- Path to the checkpoint folder.trainPath- Path to the training model.evalPath- Path to the evaluation model. Null signifies there is no eval model.optimizerPath- Path to the optimizer model. Null signifies there is no optimizer model.options- The session options.- Returns:
- An
OrtTrainingSessionwith the specified model. - Throws:
OrtException- If the model failed to load, wasn't compatible or caused an error.
-
isTrainingEnabled
public boolean isTrainingEnabled()Is training enabled in this build of ONNX Runtime?- Returns:
- True if training is enabled.
-
setTelemetry
Turns on or off the telemetry.- Parameters:
sendTelemetry- If true then send telemetry on ONNX Runtime usage.- Throws:
OrtException- If the call failed.
-
getVersion
Gets the native library version string.- Returns:
- The version string.
-
toString
-
getAvailableProviders
Gets the providers available in this environment.- Returns:
- An enum set of the available execution providers.
-
close
public void close()Close is a no-op on OrtEnvironment since ORT 1.11.- Specified by:
closein interfaceAutoCloseable
-