internal void ReadModelFromTrainedLDA(LDAModel model)
{
_ldaTrainer.GetModelStat(out var memBlockSize, out var aliasMemBlockSize);
model.MemBlockSize = memBlockSize;
model.AliasMemBlockSize = aliasMemBlockSize;
Debug.Assert(_ldaTrainer.NumVocab == model.VocabularyBuckets);
model.LDA_Data = Enumerable.Range(0, _ldaTrainer.NumVocab)
.Select(i => _ldaTrainer.GetModel(i))
.ToArray();
}
public void Save(ModelSaveContext ctx, bool saveText, VBuffer <ReadOnlyMemory <char> > mapping)
{
Contracts.AssertValue(ctx);
long memBlockSize = 0;
long aliasMemBlockSize = 0;
_ldaTrainer.GetModelStat(out memBlockSize, out aliasMemBlockSize);
// *** Binary format ***
// <ColInfoEx>
// int: vocabnum
// long: memblocksize
// long: aliasMemBlockSize
// (serializing term by term, for one term)
// int: term_id, int: topic_num, KeyValuePair<int, int>[]: termTopicVector
InfoEx.Save(ctx);
ctx.Writer.Write(_ldaTrainer.NumVocab);
ctx.Writer.Write(memBlockSize);
ctx.Writer.Write(aliasMemBlockSize);
//save model from this interface
for (int i = 0; i < _ldaTrainer.NumVocab; i++)
{
KeyValuePair <int, int>[] termTopicVector = _ldaTrainer.GetModel(i);
//write the topic to disk through ctx
ctx.Writer.Write(i); //term_id
ctx.Writer.Write(termTopicVector.Length);
foreach (KeyValuePair <int, int> p in termTopicVector)
{
ctx.Writer.Write(p.Key);
ctx.Writer.Write(p.Value);
}
}
var writeAction = GetTopicSummaryWriter(mapping);
// save word-topic summary in text
if (saveText)
{
ctx.SaveTextStream(WordTopicModelFilename, writeAction);
}
}