omega-ai-master.zip

package com.omega.engine.service.impl; import java.io.File; import org.nd4j.linalg.schedule.RampSchedule; import org.springframework.core.io.ClassPathResource; import org.springframework.core.io.Resource; import org.springframework.scheduling.annotation.Async; import org.springframework.stereotype.Service; import com.omega.common.data.Tensor; import com.omega.common.data.utils.DataTransforms; import com.omega.common.utils.DataLoader; import com.omega.common.utils.ImageUtils; import com.omega.common.utils.JsonUtils; import com.omega.common.utils.LabelUtils; import com.omega.common.utils.MathUtils; import com.omega.common.utils.RandomUtils; import com.omega.engine.controller.TrainTask; import com.omega.engine.gpu.CUDAMemoryManager; import com.omega.engine.gpu.CUDAModules; import com.omega.engine.loss.CrossEntropyLoss; import com.omega.engine.loss.LossType; import com.omega.engine.loss.SoftmaxWithCrossEntropyLoss; import com.omega.engine.nn.data.DataSet; import com.omega.engine.nn.layer.BasicBlockLayer; import com.omega.engine.nn.layer.ConvolutionLayer; import com.omega.engine.nn.layer.DropoutLayer; import com.omega.engine.nn.layer.FullyLayer; import com.omega.engine.nn.layer.InputLayer; import com.omega.engine.nn.layer.PoolingLayer; import com.omega.engine.nn.layer.SoftmaxWithCrossEntropyLayer; import com.omega.engine.nn.layer.active.LeakyReluLayer; import com.omega.engine.nn.layer.active.ReluLayer; import com.omega.engine.nn.layer.normalization.BNLayer; import com.omega.engine.nn.model.NetworkInit; import com.omega.engine.nn.network.BPNetwork; import com.omega.engine.nn.network.CNN; import com.omega.engine.optimizer.MBSGDOptimizer; import com.omega.engine.optimizer.lr.LearnRateUpdate; import com.omega.engine.pooling.PoolingType; import com.omega.engine.service.BusinessService; import com.omega.engine.updater.UpdaterType; @Service public class BusinessServiceImpl implements BusinessService { /* * @Autowired private NetworksDataBase dataBase; */ @Override public void bpNetwork_iris() { // TODO Auto-generated method stub /** * 读取训练数据集 */ String iris_train = "H:/dataset\\iris\\iris.txt"; String iris_test = "H:/dataset\\iris\\iris_test.txt"; String[] labelSet = new String[] {"1","-1"}; DataSet trainData = DataLoader.loalDataByTxt(iris_train, ",", 1, 1, 4, 2, labelSet); DataSet testData = DataLoader.loalDataByTxt(iris_test, ",", 1, 1, 4, 2, labelSet); System.out.println("train_data:"+JsonUtils.toJson(trainData)); BPNetwork netWork = new BPNetwork(new SoftmaxWithCrossEntropyLoss(), UpdaterType.adam); InputLayer inputLayer = new InputLayer(1,1,4); FullyLayer hidden1 = new FullyLayer(4, 40); ReluLayer active1 = new ReluLayer(); FullyLayer hidden2 = new FullyLayer(40, 20); ReluLayer active2 = new ReluLayer(); FullyLayer hidden3 = new FullyLayer(20, 2); SoftmaxWithCrossEntropyLayer hidden4 = new SoftmaxWithCrossEntropyLayer(2); netWork.addLayer(inputLayer); netWork.addLayer(hidden1); netWork.addLayer(active1); netWork.addLayer(hidden2); netWork.addLayer(active2); netWork.addLayer(hidden3); netWork.addLayer(hidden4); // SGDOptimizer optimizer = new SGDOptimizer(netWork, 2000, 0.001d); // BGDOptimizer optimizer = new BGDOptimizer(netWork, 20000, 0.001d); try { MBSGDOptimizer optimizer = new MBSGDOptimizer(netWork, 10, 0.00001f, 10, LearnRateUpdate.NONE, false); // netWork.GRADIENT_CHECK = true; optimizer.train(trainData); optimizer.test(testData); } catch (Exception e) { // TODO Auto-generated catch block e.printStackTrace(); }finally { try { CUDAMemoryManager.freeAll(); } catch (Exception e) { // TODO Auto-generated catch block e.printStackTrace(); } } } @Override @Async public void bpNetwork_mnist() { // TODO Auto-generated method stub /** * 读取训练数据集 */ String mnist_train_data = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\train-images.idx3-ubyte"; String mnist_train_label = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\train-labels.idx1-ubyte"; String mnist_test_data = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\t10k-images.idx3-ubyte"; String mnist_test_label = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\t10k-labels.idx1-ubyte"; String[] labelSet = new String[] {"0","1","2","3","4","5","6","7","8","9"}; DataSet trainData = DataLoader.loadDataByUByte(mnist_train_data, mnist_train_label, labelSet, 1, 1 ,784, true); DataSet testData = DataLoader.loadDataByUByte(mnist_test_data, mnist_test_label, labelSet, 1, 1 ,784, true); BPNetwork netWork = new BPNetwork(new SoftmaxWithCrossEntropyLoss(), UpdaterType.adam); netWork.learnRate = 0.001f; int inputCount = (int) (Math.sqrt(794)+10); InputLayer inputLayer = new InputLayer(1,1,784); FullyLayer hidden1 = new FullyLayer(784, inputCount, false); BNLayer bn1 = new BNLayer(); ReluLayer active1 = new ReluLayer(); FullyLayer hidden2 = new FullyLayer(inputCount, inputCount, false); BNLayer bn2 = new BNLayer(); ReluLayer active2 = new ReluLayer(); FullyLayer hidden3 = new FullyLayer(inputCount, inputCount, false); BNLayer bn3 = new BNLayer(); ReluLayer active3 = new ReluLayer(); FullyLayer hidden4 = new FullyLayer(inputCount, 10); SoftmaxWithCrossEntropyLayer softmax = new SoftmaxWithCrossEntropyLayer(10); netWork.addLayer(inputLayer); netWork.addLayer(hidden1); netWork.addLayer(bn1); netWork.addLayer(active1); netWork.addLayer(hidden2); netWork.addLayer(bn2); netWork.addLayer(active2); netWork.addLayer(hidden3); netWork.addLayer(bn3); netWork.addLayer(active3); netWork.addLayer(hidden4); netWork.addLayer(softmax); // SGDOptimizer optimizer = new SGDOptimizer(netWork, 20000, 0.001d); // BGDOptimizer optimizer = new BGDOptimizer(netWork, 20000, 0.001d); try { MBSGDOptimizer optimizer = new MBSGDOptimizer(netWork, 10, 0.001f, 128, LearnRateUpdate.NONE, false); // netWork.GRADIENT_CHECK = true; long start = System.nanoTime(); long trainTime = System.nanoTime(); optimizer.train(trainData); System.out.println("trainTime:"+((System.nanoTime() - trainTime) / 1e9) + "s."); long testTime = System.nanoTime(); optimizer.test(testData); System.out.println("testTime:"+((System.nanoTime() - testTime) / 1e9) + "s."); System.out.println(((System.nanoTime() - start) / 1e9) + "s."); } catch (Exception e) { // TODO Auto-generated catch block e.printStackTrace(); }finally { try { CUDAMemoryManager.freeAll(); } catch (Exception e) { // TODO Auto-generated catch block e.printStackTrace(); } } } @Override public void cnnNetwork_mnist_demo() { // TODO Auto-generated method stub /** * 读取训练数据集 */ String mnist_train_data = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\train-images.idx3-ubyte"; String mnist_train_label = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\train-labels.idx1-ubyte"; String mnist_test_data = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\t10k-images.idx3-ubyte"; String mnist_test_label = "C:\\Users\\Administrator\\Desktop\\dataset\\mnist\\t10k-labels.idx1-ubyte"; String[] labelSet = new String[] {"0","1","2","3","4","5","6","7","8","9"}; DataSet trainData = DataLoader.loadDataByUByte(mnist_train_data, mnist_train_label, labelSet, 1, 1 ,784,true); DataSet testData = DataLoader.loadDataByUByte(mnist_test_data, mnist_test_label, labelSet, 1, 1 ,784,true); int channel = 1; int height = 28; int width = 28; CNN netWork = new CNN(new CrossEntropyLoss(), UpdaterType.momentum); netWork.learnRate = 0.1f; InputLayer inputLayer = new InputLayer(channel, 1, 784); ConvolutionLayer conv1 = new ConvolutionLayer(channel, 6, width, height, 5, 5, 2, 1); B