yaml configuration for models, moved config out of god class, individual paramerters per model

CMakeLists.txt

@@ -24,7 +24,6 @@ if (YAMLCPP_FOUND)
   set(EXT_LIBS ${EXT_LIBS} ${YAMLCPP_LIBRARY})
 endif (YAMLCPP_FOUND)
 
-
 set(KENLM CACHE STRING "Path to compiled kenlm directory")
 if (NOT EXISTS "${KENLM}/build/lib/libkenlm.a")
   message(FATAL_ERROR "Could not find ${KENLM}/build/lib/libkenlm.a")

scripts/dropout.py

@@ -0,0 +1,30 @@
+#!/usr/bin/env python
+
+import os, sys
+import argparse
+import numpy as np;
+
+parser = argparse.ArgumentParser()
+parser.add_argument('-d', '--dropout', type=float, required=True,
+                    help="dropout rate")
+parser.add_argument('-i', '--input', required=True,
+                    help="Input model")
+parser.add_argument('-o', '--output', required=True,
+                    help="Output model")
+args = parser.parse_args()
+
+
+multiplier = 1.0 - args.dropout
+
+output = dict()
+print "Loading", args.input, "to multiple with", multiplier
+with open(args.input, "rb") as mfile:
+  m = np.load(mfile)
+  for k in m:
+    if "history_errs" in k or "_b" in k or "c_tt" in k:
+      output[k] = m[k]
+    else:
+      output[k] = multiplier * m[k]
+
+print "Saving to", args.output
+np.savez(args.output, **output)

scripts/idf.py

@@ -0,0 +1,16 @@
+import sys
+import math
+from collections import Counter
+
+c = Counter()
+N = 0
+for line in sys.stdin:
+    uniq = set(line.split())
+    for word in uniq:
+       c[word] += 1
+    N += 1
+
+keys = sorted([k for k in c])
+for word in keys:
+    idf = math.log(float(N) / float(c[word])) / math.log(N)
+    print word, ":", idf

src/CMakeLists.txt

@@ -8,6 +8,7 @@ include_directories(mblas)
 add_library(libcommon OBJECT
   cnpy/cnpy.cpp
   common/utils.cpp
+  common/exception.cpp
 )
 
 add_library(librescorer OBJECT

src/common/exception.cpp

@@ -0,0 +1,104 @@
+#include "util/exception.hh"
+
+#ifdef __GXX_RTTI
+#include <typeinfo>
+#endif
+
+#include <cerrno>
+#include <cstring>
+
+#if defined(_WIN32) || defined(_WIN64)
+#include <windows.h>
+#include <io.h>
+#endif
+
+namespace util {
+
+Exception::Exception() throw() {}
+Exception::~Exception() throw() {}
+
+void Exception::SetLocation(const char *file, unsigned int line, const char *func, const char *child_name, const char *condition) {
+  /* The child class might have set some text, but we want this to come first.
+   * Another option would be passing this information to the constructor, but
+   * then child classes would have to accept constructor arguments and pass
+   * them down.
+   */
+  std::string old_text;
+  what_.swap(old_text);
+  what_ << file << ':' << line;
+  if (func) what_ << " in " << func << " threw ";
+  if (child_name) {
+    what_ << child_name;
+  } else {
+#ifdef __GXX_RTTI
+    what_ << typeid(this).name();
+#else
+    what_ << "an exception";
+#endif
+  }
+  if (condition) {
+    what_ << " because `" << condition << '\'';
+  }
+  what_ << ".\n";
+  what_ << old_text;
+}
+
+namespace {
+
+#ifdef __GNUC__
+const char *HandleStrerror(int ret, const char *buf) __attribute__ ((unused));
+const char *HandleStrerror(const char *ret, const char * /*buf*/) __attribute__ ((unused));
+#endif
+// At least one of these functions will not be called.
+#ifdef __clang__
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wunused-function"
+#endif
+// The XOPEN version.
+const char *HandleStrerror(int ret, const char *buf) {
+  if (!ret) return buf;
+  return NULL;
+}
+
+// The GNU version.
+const char *HandleStrerror(const char *ret, const char * /*buf*/) {
+  return ret;
+}
+#ifdef __clang__
+#pragma clang diagnostic pop
+#endif
+} // namespace
+
+ErrnoException::ErrnoException() throw() : errno_(errno) {
+  char buf[200];
+  buf[0] = 0;
+#if defined(sun) || defined(_WIN32) || defined(_WIN64)
+  const char *add = strerror(errno);
+#else
+  const char *add = HandleStrerror(strerror_r(errno, buf, 200), buf);
+#endif
+
+  if (add) {
+    *this << add << ' ';
+  }
+}
+
+ErrnoException::~ErrnoException() throw() {}
+
+OverflowException::OverflowException() throw() {}
+OverflowException::~OverflowException() throw() {}
+
+#if defined(_WIN32) || defined(_WIN64)
+WindowsException::WindowsException() throw() {
+  unsigned int last_error = GetLastError();
+  char error_msg[256] = "";
+  if (!FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_IGNORE_INSERTS, NULL, last_error, LANG_NEUTRAL, error_msg, sizeof(error_msg), NULL)) {
+    *this << "Windows error " << GetLastError() << " while formatting Windows error " << last_error << ". ";
+  } else {
+    *this << "Windows error " << last_error << ": " << error_msg;
+  }
+}
+WindowsException::~WindowsException() throw() {}
+#endif
+
+} // namespace util

src/common/exception.h

@@ -0,0 +1,156 @@
+#pragma once
+
+#include "util/string_stream.hh"
+
+#include <exception>
+#include <limits>
+#include <string>
+#include <stdint.h>
+
+namespace util {
+
+template <class Except, class Data> typename Except::template ExceptionTag<Except&>::Identity operator<<(Except &e, const Data &data);
+
+class Exception : public std::exception {
+  public:
+    Exception() throw();
+    virtual ~Exception() throw();
+
+    const char *what() const throw() { return what_.str().c_str(); }
+
+    // For use by the UTIL_THROW macros.
+    void SetLocation(
+        const char *file,
+        unsigned int line,
+        const char *func,
+        const char *child_name,
+        const char *condition);
+
+  private:
+    template <class Except, class Data> friend typename Except::template ExceptionTag<Except&>::Identity operator<<(Except &e, const Data &data);
+
+    // This helps restrict operator<< defined below.
+    template <class T> struct ExceptionTag {
+      typedef T Identity;
+    };
+
+    StringStream what_;
+};
+
+/* This implements the normal operator<< for Exception and all its children.
+ * SFINAE means it only applies to Exception.  Think of this as an ersatz
+ * boost::enable_if.
+ */
+template <class Except, class Data> typename Except::template ExceptionTag<Except&>::Identity operator<<(Except &e, const Data &data) {
+  e.what_ << data;
+  return e;
+}
+
+#ifdef __GNUC__
+#define UTIL_FUNC_NAME __PRETTY_FUNCTION__
+#else
+#ifdef _WIN32
+#define UTIL_FUNC_NAME __FUNCTION__
+#else
+#define UTIL_FUNC_NAME NULL
+#endif
+#endif
+
+/* Create an instance of Exception, add the message Modify, and throw it.
+ * Modify is appended to the what() message and can contain << for ostream
+ * operations.
+ *
+ * do .. while kludge to swallow trailing ; character
+ * http://gcc.gnu.org/onlinedocs/cpp/Swallowing-the-Semicolon.html .
+ * Arg can be a constructor argument to the exception.
+ */
+#define UTIL_THROW_BACKEND(Condition, Exception, Arg, Modify) do { \
+  Exception UTIL_e Arg; \
+  UTIL_e.SetLocation(__FILE__, __LINE__, UTIL_FUNC_NAME, #Exception, Condition); \
+  UTIL_e << Modify; \
+  throw UTIL_e; \
+} while (0)
+
+#define UTIL_THROW_ARG(Exception, Arg, Modify) \
+  UTIL_THROW_BACKEND(NULL, Exception, Arg, Modify)
+
+#define UTIL_THROW(Exception, Modify) \
+  UTIL_THROW_BACKEND(NULL, Exception, , Modify);
+
+#define UTIL_THROW2(Modify) \
+  UTIL_THROW_BACKEND(NULL, util::Exception, , Modify);
+
+#if __GNUC__ >= 3
+#define UTIL_UNLIKELY(x) __builtin_expect (!!(x), 0)
+#else
+#define UTIL_UNLIKELY(x) (x)
+#endif
+
+#if __GNUC__ >= 3
+#define UTIL_LIKELY(x) __builtin_expect (!!(x), 1)
+#else
+#define UTIL_LIKELY(x) (x)
+#endif
+
+#define UTIL_THROW_IF_ARG(Condition, Exception, Arg, Modify) do { \
+  if (UTIL_UNLIKELY(Condition)) { \
+    UTIL_THROW_BACKEND(#Condition, Exception, Arg, Modify); \
+  } \
+} while (0)
+
+#define UTIL_THROW_IF(Condition, Exception, Modify) \
+  UTIL_THROW_IF_ARG(Condition, Exception, , Modify)
+
+#define UTIL_THROW_IF2(Condition, Modify) \
+  UTIL_THROW_IF_ARG(Condition, util::Exception, , Modify)
+
+// Exception that records errno and adds it to the message.
+class ErrnoException : public Exception {
+  public:
+    ErrnoException() throw();
+
+    virtual ~ErrnoException() throw();
+
+    int Error() const throw() { return errno_; }
+
+  private:
+    int errno_;
+};
+
+// file wasn't there, or couldn't be open for some reason
+class FileOpenException : public Exception {
+  public:
+	FileOpenException() throw() {}
+    ~FileOpenException() throw() {}
+};
+
+// Utilities for overflow checking.
+class OverflowException : public Exception {
+  public:
+    OverflowException() throw();
+    ~OverflowException() throw();
+};
+
+template <unsigned len> inline std::size_t CheckOverflowInternal(uint64_t value) {
+  UTIL_THROW_IF(value > static_cast<uint64_t>(std::numeric_limits<std::size_t>::max()), OverflowException, "Integer overflow detected.  This model is too big for 32-bit code.");
+  return value;
+}
+
+template <> inline std::size_t CheckOverflowInternal<8>(uint64_t value) {
+  return value;
+}
+
+inline std::size_t CheckOverflow(uint64_t value) {
+  return CheckOverflowInternal<sizeof(std::size_t)>(value);
+}
+
+#if defined(_WIN32) || defined(_WIN64)
+/* Thrown for Windows specific operations. */
+class WindowsException : public Exception {
+  public:
+    WindowsException() throw();
+    ~WindowsException() throw();
+};
+#endif
+
+} // namespace util

src/decoder/ape_penalty.h

@@ -7,7 +7,7 @@
 #include "matrix.h"
 
 class ApePenaltyState : public State {
-  // Dummy
+  // Dummy, this scorer is stateless
 };
 
 class ApePenalty : public Scorer {
@@ -17,6 +17,7 @@ class ApePenalty : public Scorer {
     : Scorer(sourceIndex)
     { }
     
+    // @TODO: make this work on GPU
     virtual void SetSource(const Sentence& source) {
       const Words& words = source.GetWords(sourceIndex_);
       const Vocab& svcb = God::GetSourceVocab(sourceIndex_);
@@ -57,3 +58,18 @@ class ApePenalty : public Scorer {
   private:
     std::vector<float> costs_;
 };
+
+class ApePenaltyLoader : public Loader {
+  public:
+    ApePenaltyLoader(const YAML::Node& config)
+     : Loader(config) {}
+  
+    virtual void Load() {
+      // @TODO: IDF weights
+    }
+  
+    virtual ScorerPtr NewScorer(size_t taskId) {
+      size_t tab = Has("tab") ? Get<size_t>("tab") : 0;
+      return ScorerPtr(new ApePenalty(tab));
+    }
+};

src/decoder/config.cpp

@@ -1,6 +1,7 @@
 #include <set>
 
 #include "config.h"
+#include "exception.h"
 
 #define SET_OPTION(key, type) \
 if(!vm_[key].defaulted() || !config_[key]) { \
@@ -26,35 +27,20 @@ void Config::AddOptions(size_t argc, char** argv) {
 
   std::string configPath;
   std::vector<size_t> devices;
-  std::vector<size_t> tabMap;
-  std::vector<float> weights;
-  
-  std::vector<std::string> modelPaths;
-  std::vector<std::string> lmPaths;
-  std::vector<std::string> sourceVocabPaths;
-  std::string targetVocabPath;
   
   general.add_options()
-    ("config,c", po::value(&configPath),
+    ("config,c", po::value(&configPath)->required(),
      "Configuration file")
-    ("model,m", po::value(&modelPaths)->multitoken()->required(),
-     "Path to neural translation model(s)")
-    ("source,s", po::value(&sourceVocabPaths)->multitoken()->required(),
-     "Path to source vocabulary file.")
-    ("target,t", po::value(&targetVocabPath)->required(),
-     "Path to target vocabulary file.")
-    ("ape", po::value<bool>()->zero_tokens()->default_value(false),
-     "Add APE-penalty")
-    ("lm,l", po::value(&lmPaths)->multitoken(),
-     "Path to KenLM language model(s)")
-    ("tab-map", po::value(&tabMap)->multitoken()->default_value(std::vector<size_t>(1, 0), "0"),
-     "tab map")
     ("devices,d", po::value(&devices)->multitoken()->default_value(std::vector<size_t>(1, 0), "0"),
      "CUDA device(s) to use, set to 0 by default, "
      "e.g. set to 0 1 to use gpu0 and gpu1. "
      "Implicitly sets minimal number of threads to number of devices.")
     ("threads-per-device", po::value<size_t>()->default_value(1),
      "Number of threads per device, total thread count equals threads x devices")
+    ("show-weights", po::value<bool>()->zero_tokens()->default_value(false),
+     "Output used weights to stdout and exit")
+    ("load-weights", po::value<std::string>(),
+     "Load scorer weights from this file")
     ("help,h", po::value<bool>()->zero_tokens()->default_value(false),
      "Print this help message and exit")
   ;
@@ -67,26 +53,11 @@ void Config::AddOptions(size_t argc, char** argv) {
      "Normalize scores by translation length after decoding")
     ("n-best", po::value<bool>()->zero_tokens()->default_value(false),
      "Output n-best list with n = beam-size")
-    ("weights,w", po::value(&weights)->multitoken()->default_value(std::vector<float>(1, 1.0), "1.0"),
-     "Model weights (for neural models and KenLM models)")
-    ("show-weights", po::value<bool>()->zero_tokens()->default_value(false),
-     "Output used weights to stdout and exit")
-    ("load-weights", po::value<std::string>(),
-     "Load scorer weights from this file")
-  ;
-
-  po::options_description kenlm("KenLM specific options");
-  kenlm.add_options()
-    ("kenlm-batch-size", po::value<size_t>()->default_value(1000),
-     "Batch size for batched queries to KenLM")
-    ("kenlm-batch-threads", po::value<size_t>()->default_value(4),
-     "Concurrent worker threads for batch processing")
   ;
 
   po::options_description cmdline_options("Allowed options");
   cmdline_options.add(general);
   cmdline_options.add(search);
-  cmdline_options.add(kenlm);
   
   po::variables_map vm_;
   try {
@@ -108,32 +79,35 @@ void Config::AddOptions(size_t argc, char** argv) {
     exit(0);
   }
   
-  if(configPath.size())
   config_ = YAML::LoadFile(configPath);
    
-  SET_OPTION("model", std::vector<std::string>)
-  SET_OPTION_NONDEFAULT("lm", std::vector<std::string>)
-  SET_OPTION("ape", bool)
-  SET_OPTION("source", std::vector<std::string>)
-  SET_OPTION("target", std::string)
-  
   SET_OPTION("n-best", bool)
   SET_OPTION("normalize", bool)
   SET_OPTION("beam-size", size_t)
   SET_OPTION("threads-per-device", size_t)
   SET_OPTION("devices", std::vector<size_t>)
-  SET_OPTION("tab-map", std::vector<size_t>)
-  
-  SET_OPTION("weights", std::vector<float>)
   SET_OPTION("show-weights", bool)
   SET_OPTION_NONDEFAULT("load-weights", std::string)
     
-  SET_OPTION("kenlm-batch-size", size_t)
-  SET_OPTION("kenlm-batch-threads", size_t)
+  Validate();
+}
+
+void Config::Validate() {
+  UTIL_THROW_IF2(!config_["scorers"] || config_["scorers"].size() == 0,
+                 "No scorers given in config file");
+  
+  UTIL_THROW_IF2(!config_["source-vocab"] || config_["source-vocab"].size() == 0,
+                 "No source-vocab given in config file");
+  
+  UTIL_THROW_IF2(!config_["target-vocab"],
+                 "No target-vocab given in config file");
+  
+  UTIL_THROW_IF2(config_["weights"].size() != config_["scorers"].size(),
+                "Different number of models and weights in config file");
 }
 
 void OutputRec(const YAML::Node node, YAML::Emitter& out) {
-  std::set<std::string> flow = { "weights", "devices", "tab-map" };
+  std::set<std::string> flow = { "weights", "devices"};
   std::set<std::string> sorter;
   switch (node.Type()) {
     case YAML::NodeType::Null:

src/decoder/config.h

@@ -27,5 +27,6 @@ class Config {
       return out;
     }
     
+    void Validate();
     void LogOptions();
 };

src/decoder/encoder_decoder.h

@@ -1,11 +1,15 @@
 #pragma once
 
 #include <vector>
+#include <yaml-cpp/yaml.h>
 
 #include "matrix.h"
 #include "scorer.h"
+#include "loader.h"
 #include "dl4mt.h"
 
+#include "threadpool.h"
+
 class EncoderDecoderState : public State {
   public:
     mblas::Matrix& GetStates() {
@@ -34,8 +38,8 @@ class EncoderDecoder : public Scorer {
     typedef EncoderDecoderState EDState;
     
   public:
-    EncoderDecoder(const Weights& model, size_t sourceIndex)
-    : Scorer(sourceIndex), model_(model),
+    EncoderDecoder(const Weights& model, size_t tabIndex)
+    : Scorer(tabIndex), model_(model),
       encoder_(new Encoder(model_)), decoder_(new Decoder(model_))
     {}
     
@@ -98,3 +102,33 @@ class EncoderDecoder : public Scorer {
     
     mblas::Matrix SourceContext_;
 };
+
+class EncoderDecoderLoader : public Loader {
+  public:
+    EncoderDecoderLoader(const YAML::Node& config)
+     : Loader(config) {}
+  
+    virtual void Load() {
+      std::string path = Get<std::string>("path");
+      auto devices = God::Get<std::vector<size_t>>("devices");
+      ThreadPool devicePool(devices.size());
+      for(auto d : devices) {
+        devicePool.enqueue([d, &path, this] {
+          LOG(info) << "Loading model " << path << " onto gpu" << d;
+          cudaSetDevice(d);
+          weights_.emplace_back(new Weights(path, d));
+        });
+      }
+    }
+  
+    virtual ScorerPtr NewScorer(size_t taskId) {
+      size_t i = taskId % weights_.size();
+      size_t d = weights_[i]->GetDevice();
+      cudaSetDevice(d);
+      size_t tab = Has("tab") ? Get<size_t>("tab") : 0;
+      return ScorerPtr(new EncoderDecoder(*weights_[i], tab));
+    }
+    
+  private:
+    std::vector<std::unique_ptr<Weights>> weights_;
+};

src/decoder/god.cu

@@ -7,9 +7,7 @@
 #include "config.h"
 #include "scorer.h"
 #include "threadpool.h"
-#include "encoder_decoder.h"
-#include "language_model.h"
-#include "ape_penalty.h"
+#include "loader_factory.h"
 
 God God::instance_;
 
@@ -27,37 +25,11 @@ God& God::NonStaticInit(int argc, char** argv) {
   config_.AddOptions(argc, argv);
   config_.LogOptions();
   
-  for(auto sourceVocabPath : Get<std::vector<std::string>>("source"))
+  for(auto sourceVocabPath : Get<std::vector<std::string>>("source-vocab"))
     sourceVocabs_.emplace_back(new Vocab(sourceVocabPath));
-  targetVocab_.reset(new Vocab(Get<std::string>("target")));
+  targetVocab_.reset(new Vocab(Get<std::string>("target-vocab")));
 
-  auto modelPaths = Get<std::vector<std::string>>("model");
-  
-  tabMap_ =  Get<std::vector<size_t>>("tab-map");
-  if(tabMap_.size() < modelPaths.size()) {
-    // this should be a warning
-    LOG(info) << "More neural models than tabs, setting missing tabs to 0";
-    tabMap_.resize(modelPaths.size(), 0);
-  }
-  
-  // @TODO: handle this better!
   weights_ = Get<std::vector<float>>("weights");
-  if(weights_.size() < modelPaths.size()) {
-    // this should be a warning
-    LOG(info) << "More neural models than weights, setting weights to 1.0";
-    weights_.resize(modelPaths.size(), 1.0);
-  }
-  
-  if(Get<bool>("ape") && weights_.size() < modelPaths.size() + 1) {
-    LOG(info) << "Adding weight for APE-penalty: " << 1.0;
-    weights_.resize(modelPaths.size(), 1.0);
-  }
-
-  //if(weights_.size() < modelPaths.size() + lmPaths.size()) {
-  //  // this should be a warning
-  //  LOG(info) << "More KenLM models than weights, setting weights to 0.0";
-  //  weights_.resize(weights_.size() + lmPaths.size(), 0.0);
-  //}
     
   if(Has("load-weights")) {
     LoadWeights(Get<std::string>("load-weights"));
@@ -71,25 +43,8 @@ God& God::NonStaticInit(int argc, char** argv) {
     exit(0);
   }
   
-  auto devices = Get<std::vector<size_t>>("devices");
-  modelsPerDevice_.resize(devices.size());
-  {
-    ThreadPool devicePool(devices.size());
-    for(auto& modelPath : modelPaths) {
-      for(size_t i = 0; i < devices.size(); ++i) {
-        devicePool.enqueue([i, &devices, &modelPath, this]{
-          LOG(info) << "Loading model " << modelPath << " onto gpu" << devices[i];
-          cudaSetDevice(devices[i]);
-          modelsPerDevice_[i].emplace_back(new Weights(modelPath, devices[i]));
-        });
-      }
-    }
-  }
-
-  //for(auto& lmPath : lmPaths) {
-  //  LOG(info) << "Loading lm " << lmPath;
-  //  lms_.emplace_back(lmPath, *targetVocab_);
-  //}
+  for(auto&& modelConfig : config_.Get()["scorers"])
+    loaders_.emplace_back(LoaderFactory::Create(modelConfig));
   
   return *this;
 }
@@ -102,18 +57,10 @@ Vocab& God::GetTargetVocab() {
   return *Summon().targetVocab_;
 }
 
-std::vector<ScorerPtr> God::GetScorers(size_t threadId) {
-  size_t deviceId = threadId % Summon().modelsPerDevice_.size();
-  size_t device = Summon().modelsPerDevice_[deviceId][0]->GetDevice();
-  cudaSetDevice(device);
+std::vector<ScorerPtr> God::GetScorers(size_t taskId) {
   std::vector<ScorerPtr> scorers;
-  size_t i = 0;
-  for(auto& m : Summon().modelsPerDevice_[deviceId])
-    scorers.emplace_back(new EncoderDecoder(*m, Summon().tabMap_[i++]));
-  if(God::Get<bool>("ape"))
-    scorers.emplace_back(new ApePenalty(Summon().tabMap_[i++]));
-  for(auto& lm : Summon().lms_)
-    scorers.emplace_back(new LanguageModel(lm));
+  for(auto&& loader : Summon().loaders_)
+    scorers.emplace_back(loader->NewScorer(taskId));
   return scorers;
 }
 
@@ -121,15 +68,10 @@ std::vector<float>& God::GetScorerWeights() {
   return Summon().weights_;
 }
 
-std::vector<size_t>& God::GetTabMap() {
-  return Summon().tabMap_;
-}
-
 // clean up cuda vectors before cuda context goes out of scope
 void God::CleanUp() {
-  for(auto& models : Summon().modelsPerDevice_)
-    for(auto& m : models)
-      m.reset(nullptr);
+  for(auto& loader : Summon().loaders_)
+     loader.reset(nullptr);
 }
 
 void God::LoadWeights(const std::string& path) {

src/decoder/god.h

@@ -4,6 +4,7 @@
 #include "config.h"
 #include "types.h"
 #include "vocab.h"
+#include "loader.h"
 #include "scorer.h"
 #include "logging.h"
 
@@ -50,16 +51,9 @@ class God {
     std::vector<std::unique_ptr<Vocab>> sourceVocabs_;
     std::unique_ptr<Vocab> targetVocab_;
     
-    typedef std::unique_ptr<Weights> Model;
-    typedef std::vector<Model> Models;
-    typedef std::vector<Models> ModelsPerDevice;
+    std::vector<LoaderPtr> loaders_;
     
-    ModelsPerDevice modelsPerDevice_;
-    std::vector<LM> lms_;
-    
-    std::vector<ScorerPtr> scorers_;
     std::vector<float> weights_;
-    std::vector<size_t> tabMap_;
     
     std::shared_ptr<spdlog::logger> info_;
     std::shared_ptr<spdlog::logger> progress_;

src/decoder/loader.h

@@ -0,0 +1,32 @@
+#pragma once
+
+#include <vector>
+#include <yaml-cpp/yaml.h>
+
+#include "scorer.h"
+
+class Loader {
+  public:
+    Loader(const YAML::Node& config)
+    : config_(YAML::Clone(config)) {}
+
+    virtual ~Loader() {};
+
+    virtual void Load() = 0;
+
+    bool Has(const std::string& key) {
+      return config_[key];
+    }
+    
+    template <typename T>
+    T Get(const std::string& key) {
+      return config_[key].as<T>();
+    }
+
+    virtual ScorerPtr NewScorer(size_t) = 0;
+
+  protected:
+    const YAML::Node config_;
+};
+
+typedef std::unique_ptr<Loader> LoaderPtr;

src/decoder/loader_factory.h

@@ -0,0 +1,46 @@
+#pragma once
+
+#include <vector>
+#include <yaml-cpp/yaml.h>
+
+#include "exception.h"
+#include "scorer.h"
+#include "encoder_decoder.h"
+#include "ape_penalty.h"
+
+#ifdef KENLM
+#include "language_model.h"
+#endif
+
+#define IF_MATCH_RETURN(typeStr, nameStr, LoaderType) \
+do { \
+  if(typeStr == nameStr) { \
+    LoaderPtr loader(new LoaderType(config)); \
+    loader->Load(); \
+    return loader; \
+  } \
+} while(0)
+
+class LoaderFactory {
+  public:
+    static LoaderPtr Create(const YAML::Node& config) {
+      UTIL_THROW_IF2(!config["type"],
+                     "Missing scorer type in config file");
+      
+      auto type = config["type"].as<std::string>();
+      IF_MATCH_RETURN(type, "Nematus", EncoderDecoderLoader);
+      IF_MATCH_RETURN(type, "nematus", EncoderDecoderLoader);
+      IF_MATCH_RETURN(type, "NEMATUS", EncoderDecoderLoader);
+      
+      IF_MATCH_RETURN(type, "Ape", ApePenaltyLoader);
+      IF_MATCH_RETURN(type, "ape", ApePenaltyLoader);
+      IF_MATCH_RETURN(type, "APE", ApePenaltyLoader);
+#ifdef KENLM
+      IF_MATCH_RETURN(type, "KenLM", KenLMLoader)
+      IF_MATCH_RETURN(type, "kenlm", KenLMLoader)
+      IF_MATCH_RETURN(type, "KENLM", KenLMLoader)
+#endif
+      UTIL_THROW2("Unknown scorer in config file: " << type);
+    }
+};
+

src/decoder/sentence.cu

@@ -5,13 +5,11 @@
 Sentence::Sentence(size_t lineNo, const std::string& line)
 : lineNo_(lineNo), line_(line)
 {
-  auto& tabMap = God::GetTabMap();
-  
   std::vector<std::string> tabs;
   Split(line, tabs, "\t");
   size_t i = 0;
   for(auto&& tab : tabs)
-    words_.push_back(God::GetSourceVocab(tabMap[i++])(tab));
+    words_.push_back(God::GetSourceVocab(i++)(tab));
 }
 
 const Words& Sentence::GetWords(size_t index) const {