From b50c6ff365185c40ee4524b2a6a5530cee7cb4a0 Mon Sep 17 00:00:00 2001
From: Franck Dary <franck.dary@lis-lab.fr>
Date: Wed, 4 Mar 2020 13:37:34 +0100
Subject: [PATCH] Made a CNN module
---
torch_modules/include/CNN.hpp | 26 +++++++++++++++++
torch_modules/include/CNNNetwork.hpp | 6 ++--
torch_modules/src/CNN.cpp | 34 ++++++++++++++++++++++
torch_modules/src/CNNNetwork.cpp | 43 ++++++----------------------
4 files changed, 72 insertions(+), 37 deletions(-)
create mode 100644 torch_modules/include/CNN.hpp
create mode 100644 torch_modules/src/CNN.cpp
diff --git a/torch_modules/include/CNN.hpp b/torch_modules/include/CNN.hpp
new file mode 100644
index 0000000..e08a869
--- /dev/null
+++ b/torch_modules/include/CNN.hpp
@@ -0,0 +1,26 @@
+#ifndef CNN__H
+#define CNN__H
+
+#include <torch/torch.h>
+#include "fmt/core.h"
+
+class CNNImpl : public torch::nn::Module
+{
+ private :
+
+ std::vector<long> windowSizes;
+ std::vector<torch::nn::Conv2d> CNNs;
+ int nbFilters;
+ int elementSize;
+
+ public :
+
+ CNNImpl(std::vector<long> windowSizes, int nbFilters, int elementSize);
+ torch::Tensor forward(torch::Tensor input);
+ int getOutputSize();
+
+};
+TORCH_MODULE(CNN);
+
+#endif
+
diff --git a/torch_modules/include/CNNNetwork.hpp b/torch_modules/include/CNNNetwork.hpp
index b6b5cef..0893ff9 100644
--- a/torch_modules/include/CNNNetwork.hpp
+++ b/torch_modules/include/CNNNetwork.hpp
@@ -2,12 +2,12 @@
#define CNNNETWORK__H
#include "NeuralNetwork.hpp"
+#include "CNN.hpp"
class CNNNetworkImpl : public NeuralNetworkImpl
{
private :
- static inline std::vector<long> windowSizes{2,3,4};
static constexpr unsigned int maxNbLetters = 10;
private :
@@ -19,8 +19,8 @@ class CNNNetworkImpl : public NeuralNetworkImpl
torch::nn::Embedding wordEmbeddings{nullptr};
torch::nn::Linear linear1{nullptr};
torch::nn::Linear linear2{nullptr};
- std::vector<torch::nn::Conv2d> CNNs;
- std::vector<torch::nn::Conv2d> lettersCNNs;
+ CNN contextCNN{nullptr};
+ CNN lettersCNN{nullptr};
public :
diff --git a/torch_modules/src/CNN.cpp b/torch_modules/src/CNN.cpp
new file mode 100644
index 0000000..f033403
--- /dev/null
+++ b/torch_modules/src/CNN.cpp
@@ -0,0 +1,34 @@
+#include "CNN.hpp"
+#include "CNN.hpp"
+
+CNNImpl::CNNImpl(std::vector<long> windowSizes, int nbFilters, int elementSize)
+ : windowSizes(windowSizes), nbFilters(nbFilters), elementSize(elementSize)
+{
+ for (auto & windowSize : windowSizes)
+ {
+ std::string moduleName = fmt::format("cnn_window_{}", windowSize);
+ CNNs.emplace_back(register_module(moduleName, torch::nn::Conv2d(torch::nn::Conv2dOptions(1, nbFilters, torch::ExpandingArray<2>({windowSize,elementSize})).padding({windowSize-1, 0}))));
+ }
+}
+
+torch::Tensor CNNImpl::forward(torch::Tensor input)
+{
+ std::vector<torch::Tensor> windows;
+ for (unsigned int i = 0; i < CNNs.size(); i++)
+ {
+ auto convOut = torch::relu(CNNs[i](input).squeeze(-1));
+ auto pooled = torch::max_pool1d(convOut, convOut.size(2));
+ windows.emplace_back(pooled);
+ }
+
+ auto cnnOut = torch::cat(windows, 2);
+ cnnOut = cnnOut.view({cnnOut.size(0), -1});
+
+ return cnnOut;
+}
+
+int CNNImpl::getOutputSize()
+{
+ return windowSizes.size()*nbFilters;
+}
+
diff --git a/torch_modules/src/CNNNetwork.cpp b/torch_modules/src/CNNNetwork.cpp
index 50f9c0d..86781c9 100644
--- a/torch_modules/src/CNNNetwork.cpp
+++ b/torch_modules/src/CNNNetwork.cpp
@@ -13,13 +13,10 @@ CNNNetworkImpl::CNNNetworkImpl(int nbOutputs, int leftBorder, int rightBorder, i
setColumns(columns);
wordEmbeddings = register_module("word_embeddings", torch::nn::Embedding(torch::nn::EmbeddingOptions(50000, embeddingsSize)));
- linear1 = register_module("linear1", torch::nn::Linear(nbFilters*windowSizes.size()+nbFiltersLetters*windowSizes.size()*(focusedBufferIndexes.size()+focusedStackIndexes.size()), hiddenSize));
+ contextCNN = register_module("contextCNN", CNN(std::vector<long>{2,3,4}, nbFilters, 2*embeddingsSize));
+ lettersCNN = register_module("lettersCNN", CNN(std::vector<long>{2,3,4,5}, nbFiltersLetters, embeddingsSize));
+ linear1 = register_module("linear1", torch::nn::Linear(contextCNN->getOutputSize()+lettersCNN->getOutputSize()*(focusedBufferIndexes.size()+focusedStackIndexes.size()), hiddenSize));
linear2 = register_module("linear2", torch::nn::Linear(hiddenSize, nbOutputs));
- for (auto & windowSize : windowSizes)
- {
- CNNs.emplace_back(register_module(fmt::format("cnn_context_{}", windowSize), torch::nn::Conv2d(torch::nn::Conv2dOptions(1, nbFilters, torch::ExpandingArray<2>({windowSize,2*embeddingsSize})).padding({windowSize-1, 0}))));
- lettersCNNs.emplace_back(register_module(fmt::format("cnn_letters_{}", windowSize), torch::nn::Conv2d(torch::nn::Conv2dOptions(1, nbFiltersLetters, torch::ExpandingArray<2>({windowSize,embeddingsSize})).padding({windowSize-1, 0}))));
- }
}
torch::Tensor CNNNetworkImpl::forward(torch::Tensor input)
@@ -34,38 +31,16 @@ torch::Tensor CNNNetworkImpl::forward(torch::Tensor input)
auto lettersEmbeddings = wordEmbeddings(wordLetters).view({wordLetters.size(0), wordLetters.size(1)/maxNbLetters, maxNbLetters, wordEmbeddings->options.embedding_dim()}).unsqueeze(1);
auto permuted = lettersEmbeddings.permute({2,0,1,3,4});
- std::vector<torch::Tensor> windows;
+ std::vector<torch::Tensor> cnnOuts;
for (unsigned int word = 0; word < focusedBufferIndexes.size(); word++)
- for (unsigned int i = 0; i < lettersCNNs.size(); i++)
- {
- auto input = permuted[word];
- auto convOut = torch::relu(lettersCNNs[i](input).squeeze(-1));
- auto pooled = torch::max_pool1d(convOut, convOut.size(2));
- windows.emplace_back(pooled);
- }
+ cnnOuts.emplace_back(lettersCNN(permuted[word]));
for (unsigned int word = 0; word < focusedStackIndexes.size(); word++)
- for (unsigned int i = 0; i < lettersCNNs.size(); i++)
- {
- auto input = permuted[focusedBufferIndexes.size()+word];
- auto convOut = torch::relu(lettersCNNs[i](input).squeeze(-1));
- auto pooled = torch::max_pool1d(convOut, convOut.size(2));
- windows.emplace_back(pooled);
- }
- auto lettersCnnOut = torch::cat(windows, 2);
- lettersCnnOut = lettersCnnOut.view({lettersCnnOut.size(0), -1});
-
- windows.clear();
- for (unsigned int i = 0; i < CNNs.size(); i++)
- {
- auto convOut = torch::relu(CNNs[i](embeddings).squeeze(-1));
- auto pooled = torch::max_pool1d(convOut, convOut.size(2));
- windows.emplace_back(pooled);
- }
+ cnnOuts.emplace_back(lettersCNN(permuted[word]));
+ auto lettersCnnOut = torch::cat(cnnOuts, 1);
- auto cnnOut = torch::cat(windows, 2);
- cnnOut = cnnOut.view({cnnOut.size(0), -1});
+ auto contextCnnOut = contextCNN(embeddings);
- auto totalInput = torch::cat({cnnOut, lettersCnnOut}, 1);
+ auto totalInput = torch::cat({contextCnnOut, lettersCnnOut}, 1);
return linear2(torch::relu(linear1(totalInput)));
}
--
GitLab