public static string SaveFile(Epoch E)
{
int err = 0;
StringMaker str = new StringMaker();
// All things that need to be saved to a file
List <Perceptron> input_layer = E.GetInputLayer();
List <List <Perceptron> > hidden_layers = E.GetHiddenLayers();
List <Perceptron> output_layers = E.GetOutputLayers();
// Then with total hidden layer count
str.Add(FloatByteConverter.FloatToStringOfBytes((float)hidden_layers.Count));
// Then hidden layer size
if (hidden_layers.Count == 0)
{
str.Add(FloatByteConverter.FloatToStringOfBytes(0f));
}
else
{
str.Add(FloatByteConverter.FloatToStringOfBytes((float)hidden_layers[0].Count));
}
// Then hidden layer weight count
for (int i = 0; i < hidden_layers.Count; i++)
{
for (int j = 0; j < hidden_layers[i].Count; j++)
{
float values_layer = hidden_layers[i][j].GetWeights().Length;
str.Add(FloatByteConverter.FloatToStringOfBytes(values_layer));
}
}
// Then output layer size
str.Add(FloatByteConverter.FloatToStringOfBytes((float)output_layers.Count));
// Then output weight count
float values = output_layers[0].GetWeights().Length;
str.Add(FloatByteConverter.FloatToStringOfBytes(values));
// Add all bias and weights of each hidden layers perceptrons
if (hidden_layers.Count > 0)
{
for (int i = 0; i < hidden_layers.Count; i++)
{
for (int j = 0; j < hidden_layers[i].Count; j++)
{
float[] weights = hidden_layers[i][j].GetWeights();
float bias = hidden_layers[i][j].GetBias();
//BIAS
str.Add(FloatByteConverter.FloatToStringOfBytes(bias));
//WEIGHTS
for (int c = 0; c < weights.Length; c++)
{
str.Add(FloatByteConverter.FloatToStringOfBytes(weights[c]));
SavingThread.m_Saving_Layer = (hidden_layers.Count + 1) - i;
SavingThread.m_Saving_Perceptron = (hidden_layers[i].Count) - j;
SavingThread.m_Saving_Weight = (weights.Length) - c;
}
}
}
}
// Add all bias and weights of each output layer perceptron
for (int i = 0; i < output_layers.Count; i++)
{
// Save Me
float[] weights = output_layers[i].GetWeights();
float bias = output_layers[i].GetBias();
//BIAS
str.Add(FloatByteConverter.FloatToStringOfBytes(bias));
//WEIGHTS
for (int c = 0; c < weights.Length; c++)
{
str.Add(FloatByteConverter.FloatToStringOfBytes(weights[c]));
SavingThread.m_Saving_Layer = 1;
SavingThread.m_Saving_Perceptron = (output_layers.Count) - i;
SavingThread.m_Saving_Weight = (weights.Length) - c;
}
}
// Add all bias and weights of each input layer perceptron
for (int i = 0; i < input_layer.Count; i++)
{
// Save Me
float weight = input_layer[i].GetWeight(0);
float bias = input_layer[i].GetBias();
//BIAS
str.Add(FloatByteConverter.FloatToStringOfBytes(bias));
//WEIGHT
str.Add(FloatByteConverter.FloatToStringOfBytes(weight));
SavingThread.m_Saving_Layer = 0;
SavingThread.m_Saving_Perceptron = (input_layer.Count) - i;
SavingThread.m_Saving_Weight = (input_layer.Count) - i;
}
return(str.GetString());
}
public static Epoch LoadFile(string Data)
{
int _ExpectedLength = 8;
// Expect 2 variables
if (Data.Length < _ExpectedLength)
{
Debug.LogError("File too small (a)");
return(null);
}
// File index
int FileIndex = 0;
// Important Data from start of file
int hidden_layers = (int)FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
Debug.Log("hidden layers: " + hidden_layers);
int hidden_layers_size = (int)FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
Debug.Log("hidden layer sizes: " + hidden_layers_size);
// Expect (hidden_layers * hidden_layers_size * 4)
// Expect 2 more variables (output size + output weight count)
_ExpectedLength += (hidden_layers * hidden_layers_size * 4) + 8;
if (Data.Length < _ExpectedLength)
{
Debug.LogError("File too small (b)");
return(null);
}
List <int> hidden_layer_weight_count = new List <int>();
for (int i = 0; i < hidden_layers; i++)
{
for (int j = 0; j < hidden_layers_size; j++)
{
hidden_layer_weight_count.Add((int)FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4)));
FileIndex += 4;
//Debug.Log("Hidden Layer: " + (i + 1) + ", Perceptron: " + (j + 1) + ", Size: " + hidden_layer_weight_count[j]);
}
}
int output_layer_size = (int)FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
Debug.Log("output layer size: " + output_layer_size);
int output_layer_weightcount = (int)FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
Debug.Log("Output Layer, Perceptron Size: " + output_layer_weightcount);
Debug.Log("~~~");
// Check if file is correct size
// Expect
for (int i = 0; i < hidden_layers; i++)
{
for (int j = 0; j < hidden_layers_size; j++)
{
_ExpectedLength += 4;
_ExpectedLength += hidden_layer_weight_count[j] * 4;
}
}
for (int i = 0; i < output_layer_size; i++)
{
_ExpectedLength += 4;
_ExpectedLength += output_layer_weightcount * 4;
}
for (int i = 0; i < TetrisBoard.GetDataSize(); i++)
{
_ExpectedLength += 8;
}
if (Data.Length < _ExpectedLength)
{
//#if UNITY_EDITOR
// string Message = "File Too Small\n";
// Message += "File Count: " + Data.Length.ToString() + "\n";
// Message += "Expected Count: " + _ExpectedLength.ToString() + "\n";
// EditorUtility.DisplayDialog("File Size", Message, "Continue");
//#endif
return(null);
}
else if (Data.Length > _ExpectedLength)
{
//#if UNITY_EDITOR
// string Message = "File Too Big\n";
// Message += "File Count: " + Data.Length.ToString() + "\n";
// Message += "Expected Count: " + _ExpectedLength.ToString() + "\n";
// EditorUtility.DisplayDialog("File Size", Message, "Continue");
//#endif
return(null);
}
//#if UNITY_EDITOR
// else
// {
// EditorUtility.DisplayDialog("File Size", "Correct Size", "Continue");
// }
//#endif
// Read Hidden Layers
List <List <Perceptron> > All_HiddenLayer = new List <List <Perceptron> >();
for (int i = 0; i < hidden_layers; i++)
{
List <Perceptron> HiddenLayer = new List <Perceptron>();
for (int j = 0; j < hidden_layers_size; j++)
{
// Get Bias
float bias = FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
//Debug.Log("Layer: " + (i + 1) + ", Perceptron: " + (j + 1) + ", Bias: " + bias);
// Get all weights
float[] weights = new float[hidden_layer_weight_count[j]];
for (int k = 0; k < hidden_layer_weight_count[j]; k++)
{
weights[k] = FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
//Debug.Log("Layer: " + (i + 1) + ", Perceptron: " + (j + 1) + ", Weight[" + (k) + "]: " + weights[k]);
}
Perceptron P = new Perceptron(weights, bias);
HiddenLayer.Add(P);
}
All_HiddenLayer.Add(HiddenLayer);
}
// Read Output Layer
List <Perceptron> OutputLayer = new List <Perceptron>();
for (int i = 0; i < output_layer_size; i++)
{
// Get Bias
float bias = FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
//Debug.Log("Output Perceptron: " + (i + 1) + ", Bias: " + bias);
// Get all weights
float[] weights = new float[output_layer_weightcount];
for (int k = 0; k < output_layer_weightcount; k++)
{
weights[k] = FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
//Debug.Log("Output Perceptron: " + (i + 1) + ", Weight[" + (k) + "]: " + weights[k]);
}
Perceptron P = new Perceptron(weights, bias);
OutputLayer.Add(P);
}
// Read Input Layer
List <Perceptron> InputLayer = new List <Perceptron>();
for (int i = 0; i < TetrisBoard.GetDataSize(); i++)
{
// Get Bias
float bias = FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
//Debug.Log("Output Perceptron: " + (i + 1) + ", Bias: " + bias);
// Get all weights
float[] weights = new float[1];
weights[0] = FloatByteConverter.CharStringOfBytesToFloat(Data.Substring(FileIndex, 4));
FileIndex += 4;
Perceptron P = new Perceptron(weights, bias);
InputLayer.Add(P);
}
// Created Epoch
Epoch NewEpoch = new Epoch(TetrisBoard.GetDataSize(), hidden_layers, hidden_layers_size, (int)Tetris.Move.TOTAL);
NewEpoch.SetHiddenLayerData(All_HiddenLayer);
NewEpoch.SetOutputLayerData(OutputLayer);
NewEpoch.SetInputLayerData(InputLayer);
return(NewEpoch);
}
