public void TestTrainingRegression_MeanSquared_L2() { float[] referenceOutput = new float[] { 2.466907E-12f, 0.2984998f, 3.949761E-08f, 0.000338129f, 4.667237E-06f }; Network network = Network.CreateNetworkFromJSON(Testing.Properties.Resources.ReferenceNetwork1JSON); Utils.TestTraining(network, ComputeDeviceFactory.CreateFallbackComputeDevice(), referenceOutput, new MeanSquaredErrorFunction(), TrainingConfig.Regularization.L2, 10.0f, 0.01f); }
public void TestOpenCLLayerCalc() { List <int> layerConfig = new List <int>(); layerConfig.Add(10); layerConfig.Add(512); layerConfig.Add(12); layerConfig.Add(3); layerConfig.Add(51); layerConfig.Add(30); Network networkReference = Network.CreateNetworkInitRandom(layerConfig.ToArray(), new SigmoidActivation()); var jsonData = networkReference.ExportToJSON(); Network networkCpuTrained = Network.CreateNetworkFromJSON(jsonData); Network networkOpenCLTrained = Network.CreateNetworkFromJSON(jsonData); ComputeDevice cpuCalculator = ComputeDeviceFactory.CreateFallbackComputeDevice(); ComputeDevice openCLCalculator = Utils.GetFirstOpenCLDevice(); float[] testInput = new float[layerConfig[0]]; var cpuTrainedOutput = networkCpuTrained.Compute(testInput, cpuCalculator); var openCLTrainedOutput = networkOpenCLTrained.Compute(testInput, openCLCalculator); Utils.CheckNetworkError(cpuTrainedOutput, openCLTrainedOutput); }
public void TestTrainingRegression_CrossEntropy() { float[] referenceOutput = new float[] { 3.45829E-11f, 0.139514f, 3.661705E-05f, 0.3912812f, 1.076772E-06f }; Network network = Network.CreateNetworkFromJSON(Testing.Properties.Resources.ReferenceNetwork1JSON); Utils.TestTraining(network, ComputeDeviceFactory.CreateFallbackComputeDevice(), referenceOutput, new CrossEntropyErrorFunction(), TrainingConfig.Regularization.None, 0.0f, 0.01f); }
public void TestTrainingRegression_CrossEntropy_L1() { float[] referenceOutput = new float[] { 2.571606E-11f, 0.1361623f, 3.719596E-05f, 0.3950554f, 8.929147E-07f }; Network network = Network.CreateNetworkFromJSON(Testing.Properties.Resources.ReferenceNetwork1JSON); Utils.TestTraining(network, ComputeDeviceFactory.CreateFallbackComputeDevice(), referenceOutput, new CrossEntropyErrorFunction(), TrainingConfig.Regularization.L1, 10.0f, 0.01f); }
public void TestTrainingRegression_CrossEntropy_L2() { float[] referenceOutput = new float[] { 5.959933E-11f, 0.1458118f, 4.22751E-05f, 0.3619123f, 1.531221E-06f }; Network network = Network.CreateNetworkFromJSON(Testing.Properties.Resources.ReferenceNetwork1JSON); Utils.TestTraining(network, ComputeDeviceFactory.CreateFallbackComputeDevice(), referenceOutput, new CrossEntropyErrorFunction(), TrainingConfig.Regularization.L2, 10.0f, 0.01f); }
public void TestTrainingRegression_MeanSquared() { float[] referenceOutput = new float[] { 1.485432E-12f, 0.3179092f, 3.043587E-08f, 0.0003209518f, 4.003032E-06f }; Network network = Network.CreateNetworkFromJSON(Testing.Properties.Resources.ReferenceNetwork1JSON); Utils.TestTraining(network, ComputeDeviceFactory.CreateFallbackComputeDevice(), referenceOutput, new MeanSquaredErrorFunction(), TrainingConfig.Regularization.None, 0.0f, 0.01f); }
public void TestTrainingRegression_MeanSquared_L1() { float[] referenceOutput = new float[] { 1.059097E-12f, 0.3169f, 3.026366E-08f, 0.0003242506f, 3.843542E-06f }; Network network = Network.CreateNetworkFromJSON(Testing.Properties.Resources.ReferenceNetwork1JSON); Utils.TestTraining(network, ComputeDeviceFactory.CreateFallbackComputeDevice(), referenceOutput, new MeanSquaredErrorFunction(), TrainingConfig.Regularization.L1, 10.0f, 0.01f); }
public void TestRandomNetworkWithNoLayers() { int[] referenceLayerConf = new int[] { 3, 4 }; var network = Network.CreateNetworkInitRandom(referenceLayerConf, new SigmoidActivation()); float[] result = network.Compute(new float[] { 0.2f, 0.4f, 0.5f }, ComputeDeviceFactory.CreateFallbackComputeDevice()); Assert.AreEqual(referenceLayerConf[referenceLayerConf.Length - 1], result.Length); }
public static ComputeDevice GetFirstOpenCLDevice() { foreach (var item in ComputeDeviceFactory.GetComputeDevices()) { if (item.GetDeviceAccessType().ToLower() == "opencl") { return(ComputeDeviceFactory.CreateComputeDevice(item)); } } return(null); }
private static void Train(ComputeDevice selectedDevice, int epochs) { List <TrainingSuite.TrainingData> trainingData = new List <TrainingSuite.TrainingData>(); for (int i = 0; i < 10000; i++) { float[] input = new float[1]; float[] desiredOutput = new float[1]; float rnd = (float)random.NextDouble(); input[0] = rnd; desiredOutput[0] = (float)Math.Sin((rnd - 0.5f) * 4.0f) * 0.5f + 0.5f; trainingData.Add(new TrainingSuite.TrainingData(input, desiredOutput)); } TrainingSuite suite = new TrainingSuite(trainingData); suite.config.epochs = epochs; suite.config.shuffleTrainingData = true; suite.config.miniBatchSize = 100; suite.config.costFunction = new CrossEntropyErrorFunction(); suite.config.regularization = TrainingConfig.Regularization.L2; suite.config.regularizationLambda = 0.01f; suite.config.learningRate = 0.01f; Console.WriteLine("Running training for {0} epochs!", epochs); Stopwatch sw = Stopwatch.StartNew(); int progress = 0; var promise = target_network.Train(suite, ComputeDeviceFactory.CreateFallbackComputeDevice()); Console.WriteLine("____________________"); while (!promise.IsReady()) { int progress_rounded = (int)(promise.GetTotalProgress() * 20); if (progress_rounded > progress) { ++progress; Console.Write("#"); } Thread.Sleep(50); } sw.Stop(); Console.WriteLine("#"); Console.WriteLine("Training finished! Elapsed={0}ms", sw.Elapsed.TotalMilliseconds); }
public void TestRandomNetwork() { int[] referenceLayerConf = new int[] { 3, 7, 5, 4 }; var network = Network.CreateNetworkInitRandom(referenceLayerConf, new SigmoidActivation()); Assert.AreEqual(3, network.GetLayerConfig()[0]); Assert.AreEqual(7, network.GetLayerConfig()[1]); Assert.AreEqual(5, network.GetLayerConfig()[2]); Assert.AreEqual(4, network.GetLayerConfig()[3]); float[] result = network.Compute(new float[] { 0.2f, 0.4f, 0.5f }, ComputeDeviceFactory.CreateFallbackComputeDevice()); Assert.AreEqual(referenceLayerConf[referenceLayerConf.Length - 1], result.Length); }
public MainWindow() { InitializeComponent(); RecreateNetwork(); var deviceList = ComputeDeviceFactory.GetComputeDevices(); foreach (var device in deviceList) { string item = device.GetDeviceAccessType() + " - " + device.GetDeviceName(); cmbComputeDevice.Items.Add(item); } cmbComputeDevice.SelectedIndex = 0; }
private void Form1_Load(object sender, EventArgs e) { List <int> layerConfig = new List <int>(new int[] { 5, 32, 32, 5 }); solver = Network.CreateNetworkInitRandom(layerConfig.ToArray(), new SigmoidActivation(), new DefaultWeightInitializer()); calculator = ComputeDeviceFactory.CreateFallbackComputeDevice(); foreach (var device in ComputeDeviceFactory.GetComputeDevices()) { string item = device.GetDeviceAccessType() + " - " + device.GetDeviceName(); comboBox1.Items.Add(item); } comboBox1.SelectedIndex = 0; }
public static void TestTraining(Network network, float[] referenceOutput, IErrorFunction errorFunc, TrainingSuite.TrainingConfig.Regularization regularization, float regularizationLambda, float learningRate) { List <int> layerConfig = new List <int>(); layerConfig.Add(5); layerConfig.Add(33); layerConfig.Add(12); layerConfig.Add(51); layerConfig.Add(5); #region Training List <TrainingSuite.TrainingData> trainingData = new List <TrainingSuite.TrainingData>(); for (int i = 0; i < 1000; i++) { float[] input = new float[layerConfig[0]]; float[] desiredOutput = new float[layerConfig[layerConfig.Count - 1]]; input[(i * 13426) % 5] = 1.0f; desiredOutput[(i * 13426) % 5] = 1.0f; trainingData.Add(new TrainingSuite.TrainingData(input, desiredOutput)); } TrainingSuite suite = new TrainingSuite(trainingData); suite.config.epochs = 2; suite.config.shuffleTrainingData = false; suite.config.miniBatchSize = 13; suite.config.costFunction = errorFunc; suite.config.regularization = regularization; suite.config.regularizationLambda = regularizationLambda; suite.config.learningRate = learningRate; var promise = network.Train(suite, ComputeDeviceFactory.CreateFallbackComputeDevice()); promise.Await(); #endregion float[] testInput = new float[] { 0.3f, 0.4f, 0.6f, 0.1f, 0.5f }; var result = network.Compute(testInput, ComputeDeviceFactory.CreateFallbackComputeDevice()); Utils.CheckNetworkError(referenceOutput, result); }
public void TestJSONExportImport() { List <int> layerConfig = new List <int>(); layerConfig.Add(5); layerConfig.Add(33); layerConfig.Add(12); layerConfig.Add(51); layerConfig.Add(5); float[] testInput = new float[] { 0.1f, 0.5f, 0.2f, 0.14f, 0.54f }; Network networkReference = Network.CreateNetworkInitRandom(layerConfig.ToArray(), new SigmoidActivation()); Network networkFromJSON = Network.CreateNetworkFromJSON(networkReference.ExportToJSON()); float[] outputRef = networkReference.Compute(testInput, ComputeDeviceFactory.CreateFallbackComputeDevice()); float[] outputJS = networkFromJSON.Compute(testInput, ComputeDeviceFactory.CreateFallbackComputeDevice()); Utils.CheckNetworkError(outputRef, outputJS, 0.00000001); }
private void Form1_Load(object sender, EventArgs e) { comboRegularization.SelectedIndex = 2; comboCostFunction.SelectedIndex = 1; calculator = ComputeDeviceFactory.CreateFallbackComputeDevice(); bitmap = new Bitmap(targetWidth, targetHeight, System.Drawing.Imaging.PixelFormat.Format16bppRgb565); bitmapDownscaled = new Bitmap(downScaleWidth, downScaleHeight, System.Drawing.Imaging.PixelFormat.Format16bppRgb565); ClearBitmap(); pictureBox1.Image = bitmap; foreach (var device in ComputeDeviceFactory.GetComputeDevices()) { string item = device.GetDeviceAccessType() + " - " + device.GetDeviceName(); comboBox1.Items.Add(item); } comboBox1.SelectedIndex = 0; trainingtimer.Interval = 300; trainingtimer.Tick += Trainingtimer_Tick; InitRandomNetwork(); }
public static void TestOpenCLTrainingWithConfig(IErrorFunction errorFunc, TrainingSuite.TrainingConfig.Regularization regularization, float regularizationLambda, float learningRate) { List <int> layerConfig = new List <int>(); layerConfig.Add(10); layerConfig.Add(512); layerConfig.Add(12); layerConfig.Add(3); layerConfig.Add(51); layerConfig.Add(30); Network networkReference = Network.CreateNetworkInitRandom(layerConfig.ToArray(), new SigmoidActivation()); var jsonData = networkReference.ExportToJSON(); Network networkCpuTrained = Network.CreateNetworkFromJSON(jsonData); Network networkOpenCLTrained = Network.CreateNetworkFromJSON(jsonData); var cpuCalculator = ComputeDeviceFactory.CreateFallbackComputeDevice(); var openCLCalculator = GetFirstOpenCLDevice(); var rnd = new Random(); List <TrainingSuite.TrainingData> trainingData = new List <TrainingSuite.TrainingData>(); for (int i = 0; i < 1000; i++) { float[] input = new float[layerConfig[0]]; float[] output = new float[layerConfig[layerConfig.Count - 1]]; var idx = rnd.Next(0, input.Length); input[rnd.Next(0, input.Length)] = 1.0f; for (int j = 0; j < 10; j++) { output[j * 3 + 0] = idx * 0.1f; output[j * 3 + 1] = 1.0f - (idx * 0.1f); output[j * 3 + 2] = idx * 0.05f; } trainingData.Add(new TrainingSuite.TrainingData(input, output)); } TrainingSuite suite = new TrainingSuite(trainingData); suite.config.epochs = 1; suite.config.shuffleTrainingData = false; suite.config.miniBatchSize = 13; suite.config.costFunction = errorFunc; suite.config.regularization = regularization; suite.config.regularizationLambda = regularizationLambda; suite.config.learningRate = learningRate; var promise1 = networkCpuTrained.Train(suite, cpuCalculator); var promise2 = networkOpenCLTrained.Train(suite, openCLCalculator); promise1.Await(); promise2.Await(); Assert.IsTrue(promise1.IsReady() && promise2.IsReady()); float[] testInput = new float[layerConfig[0]]; var cpuTrainedOutput = networkCpuTrained.Compute(testInput, cpuCalculator); var openCLTrainedOutput = networkOpenCLTrained.Compute(testInput, cpuCalculator); CheckNetworkError(cpuTrainedOutput, openCLTrainedOutput); }
public static void TestOpenCLTrainingWithConfig(IErrorFunction errorFunc, TrainingConfig.Regularization regularization, float regularizationLambda, float learningRate, bool mix_activations = false) { IActivationFunction alternateActivation = new SigmoidActivation(); if (mix_activations) { alternateActivation = new ReLUActivation(); } int input_neurons = 10; var layer_config = new List <Tuple <IActivationFunction, int> >(); layer_config.Add(new Tuple <IActivationFunction, int>(new SigmoidActivation(), 512)); layer_config.Add(new Tuple <IActivationFunction, int>(alternateActivation, 12)); layer_config.Add(new Tuple <IActivationFunction, int>(new SigmoidActivation(), 3)); layer_config.Add(new Tuple <IActivationFunction, int>(alternateActivation, 51)); layer_config.Add(new Tuple <IActivationFunction, int>(new SigmoidActivation(), 30)); Network networkReference = Network.CreateNetworkInitRandom(input_neurons, layer_config); var jsonData = networkReference.ExportToJSON(); Network networkCpuTrained = Network.CreateNetworkFromJSON(jsonData); Network networkOpenCLTrained = Network.CreateNetworkFromJSON(jsonData); var cpuCalculator = ComputeDeviceFactory.CreateFallbackComputeDevice(); var openCLCalculator = GetFirstOpenCLDevice(); var rnd = new Random(); List <TrainingSuite.TrainingData> trainingData = new List <TrainingSuite.TrainingData>(); for (int i = 0; i < 1000; i++) { float[] input = new float[input_neurons]; float[] output = new float[layer_config.Last().Item2]; var idx = rnd.Next(0, input.Length); input[rnd.Next(0, input.Length)] = 1.0f; for (int j = 0; j < 10; j++) { output[j * 3 + 0] = idx * 0.1f; output[j * 3 + 1] = 1.0f - (idx * 0.1f); output[j * 3 + 2] = idx * 0.05f; } trainingData.Add(new TrainingSuite.TrainingData(input, output)); } TrainingSuite suite = new TrainingSuite(trainingData); suite.config.epochs = 1; suite.config.shuffleTrainingData = false; suite.config.miniBatchSize = 13; suite.config.costFunction = errorFunc; suite.config.regularization = regularization; suite.config.regularizationLambda = regularizationLambda; suite.config.learningRate = learningRate; var promise1 = networkCpuTrained.Train(suite, cpuCalculator); var promise2 = networkOpenCLTrained.Train(suite, openCLCalculator); promise1.Await(); promise2.Await(); Assert.IsTrue(promise1.IsReady() && promise2.IsReady()); float[] testInput = new float[input_neurons]; var cpuTrainedOutput = networkCpuTrained.Compute(testInput, cpuCalculator); var openCLTrainedOutput = networkOpenCLTrained.Compute(testInput, cpuCalculator); ValidateFloatArray(cpuTrainedOutput, openCLTrainedOutput); }
static void Main(string[] args) { Console.WriteLine(" ### macademy test console "); ComputeDevice selectedDevice = ComputeDeviceFactory.CreateFallbackComputeDevice(); Generate(); while (true) { Console.WriteLine(""); Console.Write("> "); string rawCommand = Console.ReadLine().Trim(); var commands = rawCommand.Split(' '); if (commands.Length == 0) { continue; } var nextCommand = commands[0]; try { if (nextCommand == "exit") { break; } else if (nextCommand == "help") { Console.WriteLine("General"); Console.WriteLine(" help - Displays this help message"); Console.WriteLine(" exit - Exits the app"); Console.WriteLine(""); Console.WriteLine("Device selection"); Console.WriteLine(" devices - Displays available devices"); Console.WriteLine(" select (n) - Selectes the devices with the given id"); Console.WriteLine(" info - Displays information about the selected device"); Console.WriteLine(" quicktest - Performs a quick test on the selected device"); Console.WriteLine(" benchmark [n] - Performs a benchmark on the selected device, on the given difficulty (1-10, default: 2)"); Console.WriteLine(""); Console.WriteLine("Model training and testing"); Console.WriteLine(" generate - Generates a new network"); Console.WriteLine(" train - Trains the network"); Console.WriteLine(" eval (i) - Evaluates the output of the network to the given input"); Console.WriteLine(" export [f] - Exports the current network to the filename provided (default: 'output.json')"); Console.WriteLine(" import [f] - Imports the network from the filename provided (default: 'output.json')"); } else if (nextCommand == "devices") { var devices = ComputeDeviceFactory.GetComputeDevices(); System.Console.WriteLine(String.Format("Found a total of {0} devices!", devices.Count)); int i = 0; foreach (var dev in devices) { Console.WriteLine(String.Format((i++).ToString() + ": [{0}] {1}", dev.GetDeviceAccessType(), dev.GetDeviceName())); } } else if (nextCommand.StartsWith("select")) { if (commands.Length >= 2) { var devices = ComputeDeviceFactory.GetComputeDevices(); int selectedDeviceId = 0; if (int.TryParse(commands[1], out selectedDeviceId)) { if (selectedDeviceId < 0 || selectedDeviceId >= devices.Count) { Console.WriteLine("No such device: " + selectedDeviceId); continue; } selectedDevice = ComputeDeviceFactory.CreateComputeDevice(devices[selectedDeviceId]); Console.WriteLine("Selected device: " + selectedDeviceId + ": " + selectedDevice.GetName()); } else { Console.WriteLine("Invalid device id given!"); } } else { Console.WriteLine("No device id given!"); } } else if (nextCommand == "quicktest") { Console.WriteLine("Testing on device: " + selectedDevice.GetName()); TestDevice(selectedDevice); } else if (nextCommand == "train") { int epochs = 1; if (commands.Length >= 2) { int.TryParse(commands[1], out epochs); } Train(selectedDevice, epochs); } else if (nextCommand == "eval") { if (commands.Length < 2) { Console.WriteLine("No input given!"); continue; } List <float> input_values = new List <float>(); for (int i = 1; i < commands.Length; ++i) { float input = 0; if (!float.TryParse(commands[i], out input)) { Console.WriteLine("Invalid input given at index {0}", (i - 1)); } input_values.Add(input); } float[] result = Eval(input_values.ToArray(), selectedDevice); Console.WriteLine("Network output: [" + string.Join(", ", result) + "]"); } else if (nextCommand == "generate") { Generate(); Console.WriteLine("A new network has been generated!"); } else if (nextCommand == "benchmark") { int level = 2; if (commands.Length >= 2) { try { level = Math.Max(Math.Min(10, Int32.Parse(commands[1])), 1); } catch (System.Exception) { } } BenchmarkDevice(selectedDevice, level); } else if (nextCommand == "export") { string filename = "output.json"; if (commands.Length >= 2) { filename = commands[1]; } try { System.IO.File.WriteAllText(filename, target_network.ExportToJSON()); Console.WriteLine("Model exported to: '{0}'", filename); } catch (Exception exc) { Console.WriteLine("Error: " + exc.Message); } } else if (nextCommand == "import") { string filename = "output.json"; if (commands.Length >= 2) { filename = commands[1]; } try { string json_network = System.IO.File.ReadAllText(filename); target_network = Network.CreateNetworkFromJSON(json_network); Console.WriteLine("Model imported from: '{0}'", filename); } catch (Exception exc) { Console.WriteLine("Error: " + exc.Message); } } else if (nextCommand == "info") { if (selectedDevice != null) { Console.WriteLine("Device Name: " + selectedDevice.GetName()); Console.WriteLine("Device Access: " + selectedDevice.GetDeviceAccessMode()); Console.WriteLine("Core count: " + selectedDevice.GetDeviceCoreCount()); Console.WriteLine("Memory: " + selectedDevice.GetDeviceMemorySize()); } else { Console.WriteLine("CPU Fallback device is selected!"); } } } catch (System.Exception exc) { Console.WriteLine("An error occured when running the command! " + exc.ToString()); } } }
private void comboBox1_SelectedIndexChanged(object sender, EventArgs e) { var devices = ComputeDeviceFactory.GetComputeDevices(); calculator = ComputeDeviceFactory.CreateComputeDevice(devices[comboBox1.SelectedIndex]); }
static void Main(string[] args) { Console.WriteLine(" ### macademy test console "); ComputeDevice selectedDevice = ComputeDeviceFactory.CreateFallbackComputeDevice(); while (true) { Console.WriteLine(""); string rawCommand = Console.ReadLine().Trim(); var commands = rawCommand.Split(' '); if (commands.Length == 0) { continue; } var nextCommand = commands[0]; try { if (nextCommand == "exit") { break; } else if (nextCommand == "help") { Console.WriteLine("help - Displays this help message"); Console.WriteLine("devices - Displays available devices"); Console.WriteLine("select (n) - Selectes the devices with the given id"); Console.WriteLine("info - Displays information about the selected device"); Console.WriteLine("test - Performs a quick test on the selected device"); Console.WriteLine("exit - Exits the app"); } else if (nextCommand == "devices") { var devices = ComputeDeviceFactory.GetComputeDevices(); System.Console.WriteLine(String.Format("Found a total of {0} devices!", devices.Count)); int i = 0; foreach (var dev in devices) { Console.WriteLine(String.Format((i++).ToString() + ": [{0}] {1}", dev.GetDeviceAccessType(), dev.GetDeviceName())); } } else if (nextCommand.StartsWith("select")) { if (commands.Length >= 2) { var devices = ComputeDeviceFactory.GetComputeDevices(); int selectedDeviceId = 0; if (int.TryParse(commands[1], out selectedDeviceId)) { if (selectedDeviceId < 0 || selectedDeviceId >= devices.Count) { Console.WriteLine("No such device: " + selectedDeviceId); continue; } selectedDevice = ComputeDeviceFactory.CreateComputeDevice(devices[selectedDeviceId]); Console.WriteLine("Selected device: " + selectedDeviceId + ": " + selectedDevice.GetName()); } else { Console.WriteLine("Invalid device id given!"); } } else { Console.WriteLine("No device id given!"); } } else if (nextCommand == "test") { Console.WriteLine("Testing on device: " + selectedDevice.GetName()); TestDevice(selectedDevice); } else if (nextCommand == "info") { if (selectedDevice != null) { Console.WriteLine("Device Name: " + selectedDevice.GetName()); Console.WriteLine("Device Access: " + selectedDevice.GetDeviceAccessMode()); Console.WriteLine("Core count: " + selectedDevice.GetDeviceCoreCount()); Console.WriteLine("Memory: " + selectedDevice.GetDeviceMemorySize()); } else { Console.WriteLine("CPU Fallback device is selected!"); } } } catch (System.Exception exc) { Console.WriteLine("An error occured when running the command! " + exc.ToString()); } } }
private static ComputeDevice GetCalculator(int deviceId) { return(ComputeDeviceFactory.GetComputeDeviceById(deviceId)); }