Build status for all platforms: 
Commercial support:
This directory contains the JavaCPP Presets module for:
- ONNX Runtime 1.19.2 https://microsoft.github.io/onnxruntime/
Please refer to the parent README.md file for more detailed information about the JavaCPP Presets.
Java API documentation is available here:
Here is a simple example of ONNX Runtime ported to Java from this C++ source file:
We can use Maven 3 to download and install automatically all the class files as well as the native binaries. To run this sample code, after creating the pom.xml and CXXApiSample.java source files below, simply execute on the command line:
$ mvn compile exec:java<project>
<modelVersion>4.0.0</modelVersion>
<groupId>org.bytedeco.onnxruntime</groupId>
<artifactId>cxxapisample</artifactId>
<version>1.5.11-SNAPSHOT</version>
<properties>
<exec.mainClass>CXXApiSample</exec.mainClass>
</properties>
<dependencies>
<dependency>
<groupId>org.bytedeco</groupId>
<artifactId>onnxruntime-platform</artifactId>
<version>1.19.2-1.5.11-SNAPSHOT</version>
</dependency>
<!-- Additional dependencies required to use CUDA and cuDNN -->
<dependency>
<groupId>org.bytedeco</groupId>
<artifactId>onnxruntime-platform-gpu</artifactId>
<version>1.19.2-1.5.11-SNAPSHOT</version>
</dependency>
<!-- Additional dependencies to use bundled CUDA and cuDNN -->
<dependency>
<groupId>org.bytedeco</groupId>
<artifactId>cuda-platform-redist</artifactId>
<version>12.6-9.3-1.5.11-SNAPSHOT</version>
</dependency>
</dependencies>
<build>
<sourceDirectory>.</sourceDirectory>
</build>
</project>// Copyright(c) Microsoft Corporation.All rights reserved.
// Licensed under the MIT License.
//
import org.bytedeco.javacpp.*;
import org.bytedeco.javacpp.indexer.*;
import org.bytedeco.onnxruntime.*;
import static org.bytedeco.onnxruntime.global.onnxruntime.*;
public class CXXApiSample {
public static void main(String[] args) throws Exception {
//*************************************************************************
// initialize enviroment...one enviroment per process
// enviroment maintains thread pools and other state info
Env env = new Env(ORT_LOGGING_LEVEL_WARNING, "test");
// initialize session options if needed
SessionOptions session_options = new SessionOptions();
session_options.SetIntraOpNumThreads(1);
// If onnxruntime.dll is built with CUDA enabled, we can uncomment out this line to use CUDA for this
// session (we also need to include cuda_provider_factory.h above which defines it)
// #include "cuda_provider_factory.h"
// OrtSessionOptionsAppendExecutionProvider_CUDA(session_options.asOrtSessionOptions(), 0);
OrtSessionOptionsAppendExecutionProvider_Dnnl(session_options.asOrtSessionOptions(), 1);
// Sets graph optimization level
// Available levels are
// ORT_DISABLE_ALL -> To disable all optimizations
// ORT_ENABLE_BASIC -> To enable basic optimizations (Such as redundant node removals)
// ORT_ENABLE_EXTENDED -> To enable extended optimizations (Includes level 1 + more complex optimizations like node fusions)
// ORT_ENABLE_ALL -> To Enable All possible opitmizations
session_options.SetGraphOptimizationLevel(ORT_ENABLE_EXTENDED);
//*************************************************************************
// create session and load model into memory
// using squeezenet version 1.3
// URL = https://github.com/onnx/models/tree/master/squeezenet
String s = args.length > 0 ? args[0] : "squeezenet.onnx";
Pointer model_path = Loader.getPlatform().startsWith("windows") ? new CharPointer(s) : new BytePointer(s);
System.out.println("Using Onnxruntime C++ API");
Session session = new Session(env, model_path, session_options);
//*************************************************************************
// print model input layer (node names, types, shape etc.)
AllocatorWithDefaultOptions allocator = new AllocatorWithDefaultOptions();
// print number of model input nodes
long num_input_nodes = session.GetInputCount();
PointerPointer input_node_names = new PointerPointer(num_input_nodes);
LongPointer input_node_dims = null; // simplify... this model has only 1 input node {1, 3, 224, 224}.
// Otherwise need vector<vector<>>
System.out.println("Number of inputs = " + num_input_nodes);
// iterate over all input nodes
for (long i = 0; i < num_input_nodes; i++) {
// print input node names
BytePointer input_name = session.GetInputNameAllocated(i, new OrtAllocator(allocator.asUnownedAllocator()));
System.out.println("Input " + i + " : name=" + input_name.getString());
input_node_names.put(i, input_name);
// print input node types
TypeInfo type_info = session.GetInputTypeInfo(i);
ConstTensorTypeAndShapeInfo tensor_info = type_info.GetTensorTypeAndShapeInfo();
int type = tensor_info.GetElementType();
System.out.println("Input " + i + " : type=" + type);
// print input shapes/dims
input_node_dims = tensor_info.GetShape();
System.out.println("Input " + i + " : num_dims=" + input_node_dims.capacity());
for (long j = 0; j < input_node_dims.capacity(); j++)
System.out.println("Input " + i + " : dim " + j + "=" + input_node_dims.get(j));
}
// Results should be...
// Number of inputs = 1
// Input 0 : name = data_0
// Input 0 : type = 1
// Input 0 : num_dims = 4
// Input 0 : dim 0 = 1
// Input 0 : dim 1 = 3
// Input 0 : dim 2 = 224
// Input 0 : dim 3 = 224
//*************************************************************************
// Similar operations to get output node information.
// Use OrtSessionGetOutputCount(), OrtSessionGetOutputName()
// OrtSessionGetOutputTypeInfo() as shown above.
//*************************************************************************
// Score the model using sample data, and inspect values
long input_tensor_size = 224 * 224 * 3; // simplify ... using known dim values to calculate size
// use OrtGetTensorShapeElementCount() to get official size!
FloatPointer input_tensor_values = new FloatPointer(input_tensor_size);
PointerPointer output_node_names = new PointerPointer("softmaxout_1");
// initialize input data with values in [0.0, 1.0]
FloatIndexer idx = FloatIndexer.create(input_tensor_values);
for (long i = 0; i < input_tensor_size; i++)
idx.put(i, (float)i / (input_tensor_size + 1));
// create input tensor object from data values
MemoryInfo memory_info = MemoryInfo.CreateCpu(OrtArenaAllocator, OrtMemTypeDefault);
Value input_tensor = Value.CreateTensorFloat(memory_info.asOrtMemoryInfo(), input_tensor_values, input_tensor_size, input_node_dims, 4);
assert input_tensor.IsTensor();
// score model & input tensor, get back output tensor
ValueVector output_tensor = session.Run(new RunOptions(), input_node_names, input_tensor, 1, output_node_names, 1);
assert output_tensor.size() == 1 && output_tensor.get(0).IsTensor();
// Get pointer to output tensor float values
FloatPointer floatarr = output_tensor.get(0).GetTensorMutableDataFloat();
assert Math.abs(floatarr.get(0) - 0.000045) < 1e-6;
// score the model, and print scores for first 5 classes
for (int i = 0; i < 5; i++)
System.out.println("Score for class [" + i + "] = " + floatarr.get(i));
// Results should be as below...
// Score for class[0] = 0.000045
// Score for class[1] = 0.003846
// Score for class[2] = 0.000125
// Score for class[3] = 0.001180
// Score for class[4] = 0.001317
System.out.println("Done!");
System.exit(0);
}
}