Fixed CNNNetwork

torch_modules/src/CNNNetwork.cpp

@@ -31,19 +31,19 @@ torch::Tensor CNNNetworkImpl::forward(torch::Tensor input)
     input = input.unsqueeze(0);
 
   auto wordIndexes = input.narrow(1, 0, columns.size()*(1+leftBorder+rightBorder));
-  auto curIndex = wordIndexes.size(1);
 
-  std::vector<torch::Tensor> cnnOutputs;
+  auto elementsEmbeddings = wordEmbeddings(input.narrow(1, wordIndexes.size(1), input.size(1)-wordIndexes.size(1)));
 
+  std::vector<torch::Tensor> cnnOutputs;
+  auto curIndex = 0;
   for (unsigned int i = 0; i < focusedColumns.size(); i++)
   {
-    long nbElements = input[0][curIndex].item<long>();
-
-    curIndex++;
+    long nbElements = maxNbElements[i];
     for (unsigned int focused = 0; focused < focusedBufferIndexes.size()+focusedStackIndexes.size(); focused++)
     {
-      cnnOutputs.emplace_back(cnns[i](wordEmbeddings(input.narrow(1, curIndex, nbElements)).unsqueeze(1)));
+      auto cnnInput = elementsEmbeddings.narrow(1, curIndex, nbElements).unsqueeze(1);
       curIndex += nbElements;
+      cnnOutputs.emplace_back(cnns[i](cnnInput));
     }
   }
 
@@ -60,8 +60,8 @@ std::vector<long> CNNNetworkImpl::extractContext(Config & config, Dict & dict) c
   std::vector<long> contextIndexes = extractContextIndexes(config);
   std::vector<long> context;
 
-  for (auto & col : columns)
-    for (auto index : contextIndexes)
+  for (auto index : contextIndexes)
+    for (auto & col : columns)
       if (index == -1)
         context.push_back(dict.getIndexOrInsert(Dict::nullValueStr));
       else
@@ -71,8 +71,6 @@ std::vector<long> CNNNetworkImpl::extractContext(Config & config, Dict & dict) c
   {
     auto & col = focusedColumns[colIndex];
 
-    context.push_back(maxNbElements[colIndex]);
-
     std::vector<int> focusedIndexes;
     for (auto relIndex : focusedBufferIndexes)
     {