/// <summary>
/// 编码
/// </summary>
/// <param name="sentIndex">对话索引</param>
/// <param name="OutputSentence">输出的样句</param>
/// <param name="g"></param>
/// <param name="cost"></param>
/// <param name="encoded"></param>
private void Encode(int sentIndex, out List <string> OutputSentence, out ComputeGraph g, out double cost, ref List <WeightMatrix> encoded)
{
//var sentIndex = r.Next(0, InputSequences.Count);
var inputSentence = InputSequences[sentIndex];
var reversSentence = InputSequences[sentIndex].ToList();
reversSentence.Reverse();
OutputSentence = OutputSequences[sentIndex];
g = new ComputeGraph();
cost = 0.0;
for (int i = 0; i < inputSentence.Count; i++)
{
int ix_source = wordToIndex[inputSentence[i]].i; //顺
int ix_source2 = wordToIndex[reversSentence[i]].i; //逆
var x = g.PeekRow(Embedding, ix_source); //查询指定行数据
var eOutput = encoder.Encode(x, g);
var x2 = g.PeekRow(Embedding, ix_source2);
var eOutput2 = ReversEncoder.Encode(x2, g);
encoded.Add(g.concatColumns(eOutput, eOutput2));
}
//if (UseDropout)
//{
// encoded = g.Dropout(encoded, 0.2);
//}
}
/// <summary>
///
/// </summary>
/// <param name="OutputSentence"></param>
/// <param name="g"></param>
/// <param name="cost"></param>
/// <param name="encoded"></param>
/// <returns></returns>
private double DecodeOutput(List <string> OutputSentence, ComputeGraph g, double cost, List <WeightMatrix> encoded)
{
int ix_input = 1;
for (int i = 0; i < OutputSentence.Count + 1; i++)
{
int ix_target = 0;
if (i == OutputSentence.Count)
{
ix_target = 0;
}
else
{
ix_target = wordToIndex[OutputSentence[i]].i;
}
var x = g.PeekRow(Embedding, ix_input);
var eOutput = decoder.Decode(x, encoded, g);
if (UseDropout)
{
eOutput = g.Dropout(eOutput, 0.2);
}
var o = g.add(
g.mul(eOutput, this.Whd), this.bd);
if (UseDropout)
{
o = g.Dropout(o, 0.2);
}
var probs = g.SoftmaxWithCrossEntropy(o);
cost += -Math.Log(probs.Weight[ix_target]);
o.Gradient = probs.Weight;
o.Gradient[ix_target] -= 1;
ix_input = ix_target;
}
return(cost);
}
/// <summary>
/// 预测
/// </summary>
/// <param name="inputSeq">输入分词过的语句</param>
/// <returns></returns>
public ExecuteResult <List <string> > Predict(List <string> inputSeq)
{
ExecuteResult <List <string> > eresult = new ExecuteResult <List <string> >();
Reset();
List <string> result = new List <string>();
var G2 = new ComputeGraph(false);
//反序组
List <string> revseq = inputSeq.ToList();
revseq.Reverse();
List <WeightMatrix> encoded = new List <WeightMatrix>();
//
//Console.WriteLine($"keys>{string.Join(",", wordToIndex.Keys.ToArray())}");
for (int i = 0; i < inputSeq.Count; i++)
{
//索引
if (!wordToIndex.ContainsKey(inputSeq[i]))
{
return(eresult.SetFail($"抱歉，未能理解 \"{inputSeq[i]}\" 的含义, 请重新训练我吧!"));
//return $"抱歉，未能理解 \"{inputSeq[i]}\" 的含义, 请重新训练我吧!".Split(' ').ToList();
//return $"I'm sorry, I can't understand \"{inputSeq[i]}\" the meaning of the word, please you to retrain me!".Split(' ').ToList();
}
if (!wordToIndex.ContainsKey(revseq[i]))
{
return(eresult.SetFail($"抱歉，未能理解 \"{revseq[i]}\" 的含义, 请重新训练我吧!"));
//return $"抱歉，未能理解 \"{inputSeq[i]}\" 的含义, 请重新训练我吧!".Split(' ').ToList();
//return $"I'm sorry, I can't understand \"{revseq[i]}\" the meaning of the word, please you to retrain me!".Split(' ').ToList();
}
int ix = wordToIndex[inputSeq[i]].i;
int ix2 = wordToIndex[revseq[i]].i;
var x2 = G2.PeekRow(Embedding, ix);
var o = encoder.Encode(x2, G2);
var x3 = G2.PeekRow(Embedding, ix2);
var eOutput2 = ReversEncoder.Encode(x3, G2);
var d = G2.concatColumns(o, eOutput2);
encoded.Add(d);
}
//if (UseDropout)
//{
// for (int i = 0; i < encoded.Weight.Length; i++)
// {
// encoded.Weight[i] *= 0.2;
// }
//}
var ix_input = 1;
while (true)
{
var x = G2.PeekRow(Embedding, ix_input);
var eOutput = decoder.Decode(x, encoded, G2);
if (UseDropout)
{
for (int i = 0; i < eOutput.Weight.Length; i++)
{
eOutput.Weight[i] *= 0.2;
}
}
var o = G2.add(
G2.mul(eOutput, this.Whd), this.bd);
if (UseDropout)
{
for (int i = 0; i < o.Weight.Length; i++)
{
o.Weight[i] *= 0.2;
}
}
var probs = G2.SoftmaxWithCrossEntropy(o);
var maxv = probs.Weight[0];
var maxi = 0;
for (int i = 1; i < probs.Weight.Length; i++)
{
if (probs.Weight[i] > maxv)
{
maxv = probs.Weight[i];
maxi = i;
}
}
var pred = maxi;
if (pred == 0)
{
break; // END token predicted, break out
}
if (result.Count > max_word)
{
break;
} // something is wrong
var letter2 = indexToWord[pred].w;
result.Add(letter2);
ix_input = pred;
}
return(eresult.SetData(result).SetOk());
}