internal int FeedTrain(LDAModel model, ReadOnlySpan <int> tokenIndices, int tokenCount, ReadOnlySpan <double> frequency)
{
if (tokenCount < model.MinimumTokenCountPerDocument)
{
return(0);
}
return(_ldaTrainer.LoadDoc(tokenIndices, frequency, tokenCount, model.VocabularyBuckets));
}
public int FeedTrain(IExceptionContext ectx, ref VBuffer <Double> input)
{
Contracts.AssertValue(ectx);
// REVIEW: Input the counts to your trainer here. This
// is called multiple times.
int docSize = 0;
int termNum = 0;
for (int i = 0; i < input.Count; i++)
{
int termFreq = GetFrequency(input.Values[i]);
if (termFreq < 0)
{
// Ignore this row.
return(0);
}
if (docSize >= InfoEx.NumMaxDocToken - termFreq)
{
break;
}
// If legal then add the term.
docSize += termFreq;
termNum++;
}
// Ignore empty doc.
if (docSize == 0)
{
return(0);
}
int actualSize = 0;
if (input.IsDense)
{
actualSize = _ldaTrainer.LoadDocDense(input.Values, termNum, input.Length);
}
else
{
actualSize = _ldaTrainer.LoadDoc(input.Indices, input.Values, termNum, input.Length);
}
ectx.Assert(actualSize == 2 * docSize + 1, string.Format("The doc size are distinct. Actual: {0}, Expected: {1}", actualSize, 2 * docSize + 1));
return(actualSize);
}
