diff --git a/torch_modules/include/ContextualModule.hpp b/torch_modules/include/ContextualModule.hpp
new file mode 100644
index 0000000000000000000000000000000000000000..d7e290cc066fa2a1a842b02f702c5c717390e062
--- /dev/null
+++ b/torch_modules/include/ContextualModule.hpp
@@ -0,0 +1,39 @@
+#ifndef CONTEXTUALMODULE__H
+#define CONTEXTUALMODULE__H
+
+#include <torch/torch.h>
+#include <optional>
+#include "Submodule.hpp"
+#include "MyModule.hpp"
+#include "GRU.hpp"
+#include "LSTM.hpp"
+#include "Concat.hpp"
+
+class ContextualModuleImpl : public Submodule
+{
+ private :
+
+ torch::nn::Embedding wordEmbeddings{nullptr};
+ std::shared_ptr<MyModule> myModule{nullptr};
+ std::vector<std::string> columns;
+ std::vector<std::function<std::string(const std::string &)>> functions;
+ std::vector<std::tuple<Config::Object, int, std::optional<int>>> targets;
+ std::pair<int,int> window;
+ int inSize;
+ int outSize;
+ std::filesystem::path path;
+ std::filesystem::path w2vFile;
+
+ public :
+
+ ContextualModuleImpl(std::string name, const std::string & definition, std::filesystem::path path);
+ torch::Tensor forward(torch::Tensor input);
+ std::size_t getOutputSize() override;
+ std::size_t getInputSize() override;
+ void addToContext(std::vector<std::vector<long>> & context, const Config & config) override;
+ void registerEmbeddings() override;
+};
+TORCH_MODULE(ContextualModule);
+
+#endif
+
diff --git a/torch_modules/include/ModularNetwork.hpp b/torch_modules/include/ModularNetwork.hpp
index 3fa417ba2ffbb566f79316cdd7a9b51f5a14c6d7..9b7efaec8e94d55f745aaf4d16ab7c5f5877c811 100644
--- a/torch_modules/include/ModularNetwork.hpp
+++ b/torch_modules/include/ModularNetwork.hpp
@@ -3,6 +3,7 @@
#include "NeuralNetwork.hpp"
#include "ContextModule.hpp"
+#include "ContextualModule.hpp"
#include "RawInputModule.hpp"
#include "SplitTransModule.hpp"
#include "AppliableTransModule.hpp"
diff --git a/torch_modules/src/ContextModule.cpp b/torch_modules/src/ContextModule.cpp
index 2e13833591fbac8fe99125e122d093d6b1611971..a3129c8eef06a2cf41056571c8b50eb95180f59d 100644
--- a/torch_modules/src/ContextModule.cpp
+++ b/torch_modules/src/ContextModule.cpp
@@ -148,7 +148,7 @@ torch::Tensor ContextModuleImpl::forward(torch::Tensor input)
context = context.view({context.size(0), context.size(1)/(int)columns.size(), (int)columns.size()*context.size(2)});
- return myModule->forward(context);
+ return myModule->forward(context).reshape({input.size(0), -1});
}
void ContextModuleImpl::registerEmbeddings()
diff --git a/torch_modules/src/ContextualModule.cpp b/torch_modules/src/ContextualModule.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..d5f60788f769fa801b667b1f21815a47b53a4e34
--- /dev/null
+++ b/torch_modules/src/ContextualModule.cpp
@@ -0,0 +1,194 @@
+#include "ContextualModule.hpp"
+
+ContextualModuleImpl::ContextualModuleImpl(std::string name, const std::string & definition, std::filesystem::path path) : path(path)
+{
+ setName(name);
+
+ std::regex regex("(?:(?:\\s|\\t)*)Window\\{(.*)\\}(?:(?:\\s|\\t)*)Columns\\{(.*)\\}(?:(?:\\s|\\t)*)(\\S+)\\{(.*)\\}(?:(?:\\s|\\t)*)In\\{(.*)\\}(?:(?:\\s|\\t)*)Out\\{(.*)\\}(?:(?:\\s|\\t)*)w2v\\{(.*)\\}(?:(?:\\s|\\t)*)Targets\\{(.*)\\}(?:(?:\\s|\\t)*)");
+ if (!util::doIfNameMatch(regex, definition, [this,&definition](auto sm)
+ {
+ try
+ {
+ auto splited = util::split(sm.str(1), ' ');
+ if (splited.size() != 2)
+ util::myThrow("bad Window, expected 2 indexes");
+ window = std::make_pair(std::stoi(splited[0]), std::stoi(splited[1]));
+
+ auto funcColumns = util::split(sm.str(2), ' ');
+ columns.clear();
+ for (auto & funcCol : funcColumns)
+ {
+ functions.emplace_back() = getFunction(funcCol);
+ columns.emplace_back(util::split(funcCol, ':').back());
+ }
+
+ auto subModuleType = sm.str(3);
+ auto subModuleArguments = util::split(sm.str(4), ' ');
+
+ auto options = MyModule::ModuleOptions(true)
+ .bidirectional(std::stoi(subModuleArguments[0]))
+ .num_layers(std::stoi(subModuleArguments[1]))
+ .dropout(std::stof(subModuleArguments[2]))
+ .complete(std::stoi(subModuleArguments[3]));
+
+ for (auto & target : util::split(sm.str(8), ' '))
+ {
+ auto splited = util::split(target, '.');
+ if (splited.size() != 2 and splited.size() != 3)
+ util::myThrow(fmt::format("invalid target '{}' expected 'object.index(.childIndex)'", target));
+ targets.emplace_back(std::make_tuple(Config::str2object(splited[0]), std::stoi(splited[1]), splited.size() == 3 ? std::optional<int>(std::stoi(splited[2])) : std::optional<int>()));
+ }
+
+ inSize = std::stoi(sm.str(5));
+ outSize = std::stoi(sm.str(6));
+ if (outSize % 2)
+ util::myThrow("odd out size is not supported");
+
+ if (subModuleType == "LSTM")
+ myModule = register_module("myModule", LSTM(columns.size()*inSize, outSize, options));
+ else if (subModuleType == "GRU")
+ myModule = register_module("myModule", GRU(columns.size()*inSize, outSize, options));
+ else if (subModuleType == "Concat")
+ myModule = register_module("myModule", Concat(inSize));
+ else
+ util::myThrow(fmt::format("unknown sumodule type '{}'", subModuleType));
+
+ w2vFile = sm.str(7);
+
+ if (!w2vFile.empty())
+ {
+ getDict().loadWord2Vec(this->path / w2vFile);
+ getDict().setState(Dict::State::Closed);
+ dictSetPretrained(true);
+ }
+
+ } catch (std::exception & e) {util::myThrow(fmt::format("{} in '{}'",e.what(),definition));}
+ }))
+ util::myThrow(fmt::format("invalid definition '{}'", definition));
+}
+
+std::size_t ContextualModuleImpl::getOutputSize()
+{
+ return targets.size()*outSize;
+}
+
+std::size_t ContextualModuleImpl::getInputSize()
+{
+ return columns.size()*(2+window.second-window.first)+targets.size();
+}
+
+void ContextualModuleImpl::addToContext(std::vector<std::vector<long>> & context, const Config & config)
+{
+ auto & dict = getDict();
+ std::vector<long> contextIndexes;
+ std::vector<long> targetIndexes;
+ std::map<long,long> configIndex2ContextIndex;
+
+ contextIndexes.emplace_back(-2);
+
+ for (long i = window.first; i <= window.second; i++)
+ {
+ if (config.hasRelativeWordIndex(Config::Object::Buffer, i))
+ {
+ contextIndexes.emplace_back(config.getRelativeWordIndex(Config::Object::Buffer, i));
+ configIndex2ContextIndex[contextIndexes.back()] = contextIndexes.size()-1;
+ }
+ else
+ contextIndexes.emplace_back(-1);
+ }
+
+ for (auto & target : targets)
+ if (config.hasRelativeWordIndex(std::get<0>(target), std::get<1>(target)))
+ {
+ int baseIndex = config.getRelativeWordIndex(std::get<0>(target), std::get<1>(target));
+ if (!std::get<2>(target))
+ targetIndexes.emplace_back(baseIndex);
+ else
+ {
+ int childIndex = *std::get<2>(target);
+ auto childs = util::split(config.getAsFeature(Config::childsColName, baseIndex).get(), '|');
+ int candidate = -1;
+
+ if (childIndex >= 0 and childIndex < (int)childs.size())
+ candidate = std::stoi(childs[childIndex]);
+ else if (childIndex < 0 and ((int)childs.size())+childIndex >= 0)
+ candidate = std::stoi(childs[childs.size()+childIndex]);
+
+ targetIndexes.emplace_back(candidate);
+ }
+ }
+ else
+ targetIndexes.emplace_back(-1);
+
+ for (auto index : contextIndexes)
+ for (unsigned int colIndex = 0; colIndex < columns.size(); colIndex++)
+ {
+ auto & col = columns[colIndex];
+ if (index == -1)
+ {
+ for (auto & contextElement : context)
+ contextElement.push_back(dict.getIndexOrInsert(fmt::format("{}({})", col, Dict::nullValueStr)));
+ }
+ else if (index == -2)
+ {
+ for (auto & contextElement : context)
+ contextElement.push_back(dict.getIndexOrInsert(fmt::format("{}({})", col, "_NONE_")));
+ }
+ else
+ {
+ int dictIndex;
+ if (col == Config::idColName)
+ {
+ std::string value;
+ if (config.isCommentPredicted(index))
+ value = "ID(comment)";
+ else if (config.isMultiwordPredicted(index))
+ value = "ID(multiword)";
+ else if (config.isTokenPredicted(index))
+ value = "ID(token)";
+ dictIndex = dict.getIndexOrInsert(value);
+ }
+ else if (col == Config::EOSColName)
+ {
+ dictIndex = dict.getIndexOrInsert(fmt::format("EOS({})", config.getAsFeature(col, index)));
+ }
+ else
+ dictIndex = dict.getIndexOrInsert(functions[colIndex](config.getAsFeature(col, index)));
+
+ for (auto & contextElement : context)
+ contextElement.push_back(dictIndex);
+ }
+ }
+
+ for (auto index : targetIndexes)
+ {
+ if (configIndex2ContextIndex.count(index))
+ for (auto & contextElement : context)
+ contextElement.push_back(configIndex2ContextIndex.at(index)+1);
+ else
+ for (auto & contextElement : context)
+ contextElement.push_back(0);
+ }
+}
+
+torch::Tensor ContextualModuleImpl::forward(torch::Tensor input)
+{
+ auto context = wordEmbeddings(input.narrow(1, firstInputIndex, getInputSize()-targets.size())).view({input.size(0), (2+window.second-window.first), -1});
+ auto focusedIndexes = input.narrow(1, firstInputIndex+getInputSize()-targets.size(), targets.size());
+
+ auto out = myModule->forward(context);
+
+ std::vector<torch::Tensor> batchElems;
+
+ for (unsigned int i = 0; i < input.size(0); i++)
+ batchElems.emplace_back(torch::index_select(out[i], 0, focusedIndexes[i]).view({-1}));
+
+ return torch::stack(batchElems);
+}
+
+void ContextualModuleImpl::registerEmbeddings()
+{
+ wordEmbeddings = register_module("embeddings", torch::nn::Embedding(torch::nn::EmbeddingOptions(getDict().size(), inSize)));
+ loadPretrainedW2vEmbeddings(wordEmbeddings, w2vFile.empty() ? "" : path / w2vFile);
+}
+
diff --git a/torch_modules/src/DistanceModule.cpp b/torch_modules/src/DistanceModule.cpp
index 1b8034b51e2b19ec65294e5ae8f2bc7d50cfa016..40098bc37e96c2008cf4f331b761debfaaf022f9 100644
--- a/torch_modules/src/DistanceModule.cpp
+++ b/torch_modules/src/DistanceModule.cpp
@@ -50,7 +50,7 @@ DistanceModuleImpl::DistanceModuleImpl(std::string name, const std::string & def
torch::Tensor DistanceModuleImpl::forward(torch::Tensor input)
{
- return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, getInputSize())));
+ return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, getInputSize()))).reshape({input.size(0), -1});
}
std::size_t DistanceModuleImpl::getOutputSize()
diff --git a/torch_modules/src/FocusedColumnModule.cpp b/torch_modules/src/FocusedColumnModule.cpp
index 91d22b03d3f59022487be82158a59c08d4b6fbdb..62c1aef1e364b2616ef48d6559468d944546115e 100644
--- a/torch_modules/src/FocusedColumnModule.cpp
+++ b/torch_modules/src/FocusedColumnModule.cpp
@@ -48,7 +48,7 @@ torch::Tensor FocusedColumnModuleImpl::forward(torch::Tensor input)
{
std::vector<torch::Tensor> outputs;
for (unsigned int i = 0; i < focusedBuffer.size()+focusedStack.size(); i++)
- outputs.emplace_back(myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex+i*maxNbElements, maxNbElements))));
+ outputs.emplace_back(myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex+i*maxNbElements, maxNbElements))).reshape({input.size(0), -1}));
return torch::cat(outputs, 1);
}
diff --git a/torch_modules/src/GRU.cpp b/torch_modules/src/GRU.cpp
index fa6de5ccf0d594166f2544465dc5903a204692d4..a0ab6f22cd88f37b8d6c37d7fac730c850fe7cf3 100644
--- a/torch_modules/src/GRU.cpp
+++ b/torch_modules/src/GRU.cpp
@@ -2,7 +2,7 @@
GRUImpl::GRUImpl(int inputSize, int outputSize, ModuleOptions options) : outputAll(std::get<4>(options))
{
- auto gruOptions = torch::nn::GRUOptions(inputSize, outputSize)
+ auto gruOptions = torch::nn::GRUOptions(inputSize, std::get<1>(options) ? outputSize/2 : outputSize)
.batch_first(std::get<0>(options))
.bidirectional(std::get<1>(options))
.num_layers(std::get<2>(options))
@@ -13,15 +13,7 @@ GRUImpl::GRUImpl(int inputSize, int outputSize, ModuleOptions options) : outputA
torch::Tensor GRUImpl::forward(torch::Tensor input)
{
- auto gruOut = std::get<0>(gru(input));
-
- if (outputAll)
- return gruOut.reshape({gruOut.size(0), -1});
-
- if (gru->options.bidirectional())
- return torch::cat({gruOut.narrow(1,0,1).squeeze(1), gruOut.narrow(1,gruOut.size(1)-1,1).squeeze(1)}, 1);
-
- return gruOut.narrow(1,gruOut.size(1)-1,1).squeeze(1);
+ return std::get<0>(gru(input));
}
int GRUImpl::getOutputSize(int sequenceLength)
diff --git a/torch_modules/src/HistoryModule.cpp b/torch_modules/src/HistoryModule.cpp
index c326f52ef7645dea75feed3f08b1756be50d8df2..3e09b0aa7760f687b2d70efba5e696f93a45304b 100644
--- a/torch_modules/src/HistoryModule.cpp
+++ b/torch_modules/src/HistoryModule.cpp
@@ -38,7 +38,7 @@ HistoryModuleImpl::HistoryModuleImpl(std::string name, const std::string & defin
torch::Tensor HistoryModuleImpl::forward(torch::Tensor input)
{
- return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, maxNbElements)));
+ return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, maxNbElements))).reshape({input.size(0), -1});
}
std::size_t HistoryModuleImpl::getOutputSize()
diff --git a/torch_modules/src/LSTM.cpp b/torch_modules/src/LSTM.cpp
index 2844b17a256bac5de90017343fc4c7b2ad466e89..d84af461c789d1f295640d26bfddd18208f1c89f 100644
--- a/torch_modules/src/LSTM.cpp
+++ b/torch_modules/src/LSTM.cpp
@@ -2,7 +2,7 @@
LSTMImpl::LSTMImpl(int inputSize, int outputSize, ModuleOptions options) : outputAll(std::get<4>(options))
{
- auto lstmOptions = torch::nn::LSTMOptions(inputSize, outputSize)
+ auto lstmOptions = torch::nn::LSTMOptions(inputSize, std::get<1>(options) ? outputSize/2 : outputSize)
.batch_first(std::get<0>(options))
.bidirectional(std::get<1>(options))
.num_layers(std::get<2>(options))
@@ -13,15 +13,7 @@ LSTMImpl::LSTMImpl(int inputSize, int outputSize, ModuleOptions options) : outpu
torch::Tensor LSTMImpl::forward(torch::Tensor input)
{
- auto lstmOut = std::get<0>(lstm(input));
-
- if (outputAll)
- return lstmOut.reshape({lstmOut.size(0), -1});
-
- if (lstm->options.bidirectional())
- return torch::cat({lstmOut.narrow(1,0,1).squeeze(1), lstmOut.narrow(1,lstmOut.size(1)-1,1).squeeze(1)}, 1);
-
- return lstmOut.narrow(1,lstmOut.size(1)-1,1).squeeze(1);
+ return std::get<0>(lstm(input));
}
int LSTMImpl::getOutputSize(int sequenceLength)
diff --git a/torch_modules/src/ModularNetwork.cpp b/torch_modules/src/ModularNetwork.cpp
index 685060ff36557645c50c75bf1f049e71ceb38f4f..7dcf1c5d4fe2231b479131adf9d78ce5bd1fccc4 100644
--- a/torch_modules/src/ModularNetwork.cpp
+++ b/torch_modules/src/ModularNetwork.cpp
@@ -29,6 +29,8 @@ ModularNetworkImpl::ModularNetworkImpl(std::string name, std::map<std::string,st
std::string nameH = fmt::format("{}_{}", getName(), name);
if (splited.first == "Context")
modules.emplace_back(register_module(name, ContextModule(nameH, splited.second, path)));
+ else if (splited.first == "Contextual")
+ modules.emplace_back(register_module(name, ContextualModule(nameH, splited.second, path)));
else if (splited.first == "StateName")
modules.emplace_back(register_module(name, StateNameModule(nameH, splited.second)));
else if (splited.first == "History")
diff --git a/torch_modules/src/NumericColumnModule.cpp b/torch_modules/src/NumericColumnModule.cpp
index 182502324c727217ed498709b3c0ee831bb436e6..b535ded6b006b9249ddc1574c3aae816bb36a050 100644
--- a/torch_modules/src/NumericColumnModule.cpp
+++ b/torch_modules/src/NumericColumnModule.cpp
@@ -45,7 +45,7 @@ torch::Tensor NumericColumnModuleImpl::forward(torch::Tensor input)
{
auto context = input.narrow(1, firstInputIndex, getInputSize());
auto values = torch::from_blob(context.data_ptr(), context.sizes(), context.strides(), torch::TensorOptions(torch::kDouble).requires_grad(false).device(NeuralNetworkImpl::device)).to(torch::kFloat).unsqueeze(-1).clone();
- return myModule->forward(values);
+ return myModule->forward(values).reshape({input.size(0), -1});
}
std::size_t NumericColumnModuleImpl::getOutputSize()
diff --git a/torch_modules/src/RawInputModule.cpp b/torch_modules/src/RawInputModule.cpp
index c99c4ae24f6d9ea828fe6e7de4dc4a3637753f15..8f43a2fd310853793e93caaebc348ab1ff18b0be 100644
--- a/torch_modules/src/RawInputModule.cpp
+++ b/torch_modules/src/RawInputModule.cpp
@@ -39,7 +39,7 @@ RawInputModuleImpl::RawInputModuleImpl(std::string name, const std::string & def
torch::Tensor RawInputModuleImpl::forward(torch::Tensor input)
{
- return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, getInputSize())));
+ return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, getInputSize()))).reshape({input.size(0), -1});
}
std::size_t RawInputModuleImpl::getOutputSize()
diff --git a/torch_modules/src/SplitTransModule.cpp b/torch_modules/src/SplitTransModule.cpp
index 822969f321d15dd99408a8472b21c7b289ad981a..d4f6d84067329448d1c5b9b5f6ccba01157b279f 100644
--- a/torch_modules/src/SplitTransModule.cpp
+++ b/torch_modules/src/SplitTransModule.cpp
@@ -38,7 +38,7 @@ SplitTransModuleImpl::SplitTransModuleImpl(std::string name, int maxNbTrans, con
torch::Tensor SplitTransModuleImpl::forward(torch::Tensor input)
{
- return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, getInputSize())));
+ return myModule->forward(wordEmbeddings(input.narrow(1, firstInputIndex, getInputSize()))).reshape({input.size(0), -1});
}
std::size_t SplitTransModuleImpl::getOutputSize()
diff --git a/torch_modules/src/UppercaseRateModule.cpp b/torch_modules/src/UppercaseRateModule.cpp
index 818db8b01ae2575df8c0f4b7889214a8b2fdcc5a..0452eb8db781b8e83a1e62069b88c790b1214678 100644
--- a/torch_modules/src/UppercaseRateModule.cpp
+++ b/torch_modules/src/UppercaseRateModule.cpp
@@ -43,7 +43,7 @@ torch::Tensor UppercaseRateModuleImpl::forward(torch::Tensor input)
{
auto context = input.narrow(1, firstInputIndex, getInputSize());
auto values = torch::from_blob(context.data_ptr(), context.sizes(), context.strides(), torch::TensorOptions(torch::kDouble).requires_grad(false).device(NeuralNetworkImpl::device)).to(torch::kFloat).unsqueeze(-1).clone();
- return myModule->forward(values);
+ return myModule->forward(values).reshape({input.size(0), -1});
}
std::size_t UppercaseRateModuleImpl::getOutputSize()