Merge branch 'master' into pmaster

CHANGELOG.md

@@ -12,8 +12,10 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 - Fused inplace-dropout in FFN layer in Transformer
 - `--force-decode` option for marian-decoder
 - `--output-sampling` now works with ensembles (requires proper normalization via e.g `--weights 0.5 0.5`)
+- `--valid-reset-all` option
 
 ### Fixed
+- Make concat factors not break old vector implementation
 - Use allocator in hashing
 - Read/restore checkpoints from main process only when training with MPI
 - Multi-loss casts type to first loss-type before accumulation (aborted before due to missing cast)

VERSION

@@ -1 +1 @@
-v1.11.13
+v1.11.15

azure-pipelines.yml

@@ -16,13 +16,30 @@ parameters:
 # The pipeline CI trigger is set on the branch master only and PR trigger on a
 # (non-draft) pull request to any branch
 trigger:
-- master
+  # This minimizes the number of parallel pipeline runs. When a pipeline is
+  # running, the CI waits until it is completed before starting another one.
+  batch: true
+  branches:
+    include:
+    - master
+  paths:
+    exclude:
+    - azure-regression-tests.yml
+    - contrib
+    - doc
+    - examples
+    - regression-tests
+    - scripts
+    - VERSION
+    - vs
+    - '**/*.md'
+    - '**/*.txt'
 
 pool:
   name: Azure Pipelines
 
 variables:
-  - group: marian-prod-tests
+  - group: marian-regression-tests
   - name: BOOST_ROOT_WINDOWS
     value: "C:/hostedtoolcache/windows/Boost/1.72.0/x86_64"
   - name: BOOST_URL
@@ -32,7 +49,7 @@ variables:
   - name: MKL_DIR
     value: "$(Build.SourcesDirectory)/mkl"
   - name: MKL_URL
-    value: "https://romang.blob.core.windows.net/mariandev/ci/mkl-2020.1-windows-static.zip"
+    value: "https://data.statmt.org/romang/marian-regression-tests/ci/mkl-2020.1-windows-static.zip"
   - name: VCPKG_COMMIT
     value: 2022.03.10
   - name: VCPKG_DIR
@@ -52,6 +69,7 @@ stages:
 
   ######################################################################
   - job: BuildWindows
+    cancelTimeoutInMinutes: 1
     condition: eq(${{ parameters.runBuilds }}, true)
     displayName: Windows
 
@@ -188,6 +206,7 @@ stages:
 
   ######################################################################
   - job: BuildUbuntu
+    cancelTimeoutInMinutes: 1
     condition: eq(${{ parameters.runBuilds }}, true)
     displayName: Ubuntu
     timeoutInMinutes: 120
@@ -324,6 +343,7 @@ stages:
 
   ######################################################################
   - job: BuildMacOS
+    cancelTimeoutInMinutes: 1
     condition: eq(${{ parameters.runBuilds }}, true)
     displayName: macOS CPU clang
 
@@ -373,6 +393,7 @@ stages:
 
   ######################################################################
   - job: BuildInstall
+    cancelTimeoutInMinutes: 1
     condition: eq(${{ parameters.runBuilds }}, true)
     displayName: Linux CPU library install
 
@@ -435,6 +456,7 @@ stages:
 
   ######################################################################
   - job: TestWindows
+    cancelTimeoutInMinutes: 1
     displayName: Windows CPU+FBGEMM
 
     pool:
@@ -528,14 +550,14 @@ stages:
       displayName: Machine statistics
       workingDirectory: marian-prod-tests
 
-    # The current SAS token will expire on 8/30/2023 and a new one will need to be set in Marian > Pipelines > Library
+    # The current SAS token will expire on 12/31/2023 and a new one will need to be set in Marian > Pipelines > Library
     - bash: |
         cd models
         bash download-models.sh
         ls
       displayName: Prepare tests
       env:
-          AWS_SECRET_SAS_TOKEN: $(blob-sas-token)
+        AZURE_STORAGE_SAS_TOKEN: $(marian-prod-tests-blob-sas-token)
       workingDirectory: marian-prod-tests
 
     # Avoid using $(Build.SourcesDirectory) in bash tasks because on Windows pools it uses '\'
@@ -560,6 +582,7 @@ stages:
 
   ######################################################################
   - job: TestLinux
+    cancelTimeoutInMinutes: 1
     displayName: Linux CPU+FBGEMM
 
     pool:
@@ -572,7 +595,10 @@ stages:
 
     # The following packages are already installed on Azure-hosted runners: build-essential openssl libssl-dev
     # No need to install libprotobuf{17,10,9v5} on Ubuntu {20,18,16}.04 because it is installed together with libprotobuf-dev
-    - bash: sudo apt-get install -y libgoogle-perftools-dev libprotobuf-dev protobuf-compiler gcc-9 g++-9
+    # Installing libunwind-dev fixes a bug in 2204 (the libunwind-14 and libunwind-dev conflict)
+    - bash: |
+        sudo apt-get install -y libunwind-dev
+        sudo apt-get install -y libgoogle-perftools-dev libprotobuf-dev protobuf-compiler gcc-9 g++-9
       displayName: Install packages
 
     # https://software.intel.com/content/www/us/en/develop/articles/installing-intel-free-libs-and-python-apt-repo.html
@@ -629,14 +655,14 @@ stages:
       displayName: Machine statistics
       workingDirectory: marian-prod-tests
 
-    # The current SAS token will expire on 8/30/2023 and a new one will need to be set in Marian > Pipelines > Library
+    # The current SAS token will expire on 12/31/2023 and a new one will need to be set in Marian > Pipelines > Library
     - bash: |
         cd models
         bash download-models.sh
         ls
       displayName: Prepare tests
       env:
-          AWS_SECRET_SAS_TOKEN: $(blob-sas-token)
+        AZURE_STORAGE_SAS_TOKEN: $(marian-prod-tests-blob-sas-token)
       workingDirectory: marian-prod-tests
 
     - bash: MARIAN=../marian-dev/build bash ./run_mrt.sh '#cpu' '#basics' '#devops'

azure-regression-tests.yml

@@ -14,12 +14,16 @@ trigger: none
 # Hosted Azure DevOps Pool determining OS, CUDA version and available GPUs
 pool: mariandevops-pool-m60-eus
 
+variables:
+  - group: marian-regression-tests
+
 stages:
 - stage: TestsGPU
   jobs:
 
   ######################################################################
   - job: TestsGPULinux
+    cancelTimeoutInMinutes: 1
     displayName: Linux GPU tests
     timeoutInMinutes: 120
 
@@ -103,11 +107,14 @@ stages:
       workingDirectory: build
 
     # Always run regression tests from the master branch
+    # The current SAS token will expire on 12/31/2023 and a new one will need to be set in Marian > Pipelines > Library
     - bash: |
         git checkout master
         git pull origin master
         make install
       displayName: Prepare regression tests
+      env:
+        AZURE_STORAGE_SAS_TOKEN: $(marian-pub-tests-blob-sas-token)
       workingDirectory: regression-tests
 
     # Continue on error to be able to collect outputs and publish them as an artifact

regression-tests

@@ -1 +1 @@
-Subproject commit 488d454a0177ef300eab91ab813e485d420dc38d
+Subproject commit 2a8bed3f0e937a9de2d6fa92dee3bcf482d3d47b

scripts/contrib/model_info.py

@@ -42,11 +42,15 @@ def main():
             else:
                 print(model[args.key])
         else:
+            total_nb_of_parameters = 0
             for key in model:
-                if args.matrix_shapes:
-                    print(key, model[key].shape)
+                if not key == S2S_SPECIAL_NODE:
+                    total_nb_of_parameters += np.prod(model[key].shape)
+                if args.matrix_info:
+                    print(key, model[key].shape, model[key].dtype)
                 else:
                     print(key)
+            print('Total number of parameters:', total_nb_of_parameters)
 
 
 def parse_args():
@@ -57,8 +61,8 @@ def parse_args():
                         help="print values from special:model.yml node")
     parser.add_argument("-f", "--full-matrix", action="store_true",
                         help="force numpy to print full arrays for single key")
-    parser.add_argument("-ms", "--matrix-shapes", action="store_true",
-                        help="print shapes of all arrays in the model")
+    parser.add_argument("-mi", "--matrix-info", action="store_true",
+                        help="print full matrix info for all keys. Includes shape and dtype")
     return parser.parse_args()
 
 

src/common/config_parser.cpp

@@ -269,7 +269,7 @@ void ConfigParser::addOptionsModel(cli::CLIWrapper& cli) {
       "Pool encoder states instead of using cross attention (selects first encoder state, best used with special token)");
   cli.add<int>("--transformer-dim-ffn",
       "Size of position-wise feed-forward network (transformer)",
-      2048);  
+      2048);
   cli.add<int>("--transformer-decoder-dim-ffn",
       "Size of position-wise feed-forward network in decoder (transformer). Uses --transformer-dim-ffn if 0.",
       0);
@@ -591,7 +591,9 @@ void ConfigParser::addOptionsValidation(cli::CLIWrapper& cli) {
       "Multiple metrics can be specified",
       {"cross-entropy"});
   cli.add<bool>("--valid-reset-stalled",
-     "Reset all stalled validation metrics when the training is restarted");
+     "Reset stalled validation metrics when the training is restarted");
+  cli.add<bool>("--valid-reset-all",
+     "Reset all validation metrics when the training is restarted");
   cli.add<size_t>("--early-stopping",
      "Stop if the first validation metric does not improve for arg consecutive validation steps",
      10);

src/data/corpus.cpp

@@ -128,7 +128,7 @@ SentenceTuple Corpus::next() {
           size_t vocabId = i - shift;
           bool altered;
           preprocessLine(fields[i], vocabId, curId, /*out=*/altered);
-          if (altered)
+          if(altered)
             tup.markAltered();
           addWordsToSentenceTuple(fields[i], vocabId, tup);
         }

src/data/corpus_base.cpp

@@ -476,7 +476,10 @@ void CorpusBase::addAlignmentsToBatch(Ptr<CorpusBatch> batch,
     // If the batch vector is altered within marian by, for example, case augmentation,
     // the guided alignments we received for this tuple cease to be valid.
     // Hence skip setting alignments for that sentence tuple..
-    if (!batchVector[b].isAltered()) {
+    if (batchVector[b].isAltered()) {
+      LOG_ONCE(info, "Using guided-alignment with case-augmentation is not recommended and can result in unexpected behavior");
+      aligns.push_back(WordAlignment());
+    } else {
       aligns.push_back(std::move(batchVector[b].getAlignment()));
     }
   }

src/data/corpus_base.h

@@ -56,12 +56,16 @@ public:
    * @brief Returns whether this Tuple was altered or augmented from what
    * was provided to Marian in input.
    */
-  bool isAltered() const { return altered_; }
+  bool isAltered() const { 
+    return altered_; 
+  }
 
   /**
    * @brief Mark that this Tuple was internally altered or augmented by Marian
    */
-  void markAltered() { altered_ = true; }
+  void markAltered() { 
+    altered_ = true; 
+  }
 
   /**
    * @brief Adds a new sentence at the end of the tuple.

src/data/factored_vocab.cpp

@@ -21,8 +21,7 @@ namespace marian {
   maxSizeUnused;
   // If model has already been loaded, then assume this is a shared object, and skip loading it again.
   // This can be multi-threaded, so must run under lock.
-  static std::mutex s_mtx;
-  std::lock_guard<std::mutex> criticalSection(s_mtx);
+  std::lock_guard<std::mutex> criticalSection(loadMtx_);
   if (size() != 0) {
     //LOG(info, "[vocab] Attempting to load model a second time; skipping (assuming shared vocab)");
     return size();

src/data/factored_vocab.h

@@ -110,6 +110,7 @@ private:
   Word unkId_{};
   WordLUT vocab_;
   size_t lemmaSize_;
+  std::mutex loadMtx_;
 
   // factors
   char factorSeparator_ = '|';                         // separator symbol for parsing factored words

src/graph/expression_graph.cpp

@@ -64,6 +64,14 @@ Expr ExpressionGraph::add(Expr node) {
   }
 }
 
+/**
+ * Removes the node from the set of roots (will not be initialized during back propagation)
+ * @param node a pointer to a expression node
+ */
+void ExpressionGraph::removeAsRoot(Expr node) {
+  topNodes_.erase(node);
+}
+
 // Call on every checkpoint in backwards order
 void createSubtape(Expr node) {
   auto subtape = New<std::list<Expr>>();

src/graph/expression_graph.h

@@ -676,6 +676,12 @@ public:
    * @param node a pointer to a expression node
    */
   Expr add(Expr node);
+  
+  /**
+   * Removes the node from the set of roots (will not be initialized during back propagation)
+   * @param node a pointer to a expression node
+   */
+  void removeAsRoot(Expr node);
 
   /**
    * Allocate memory for the forward pass of the given node.

src/graph/expression_operators.cpp

@@ -27,6 +27,11 @@ Expr checkpoint(Expr a) {
   return a;
 }
 
+Expr removeAsRoot(Expr a) {
+  a->graph()->removeAsRoot(a); // ugly, hence why hidden here
+  return a;
+}
+
 Expr lambda(const std::vector<Expr>& nodes, Shape shape, Type type,
             LambdaNodeFunctor fwd, size_t hash) {
   return Expression<LambdaNodeOp>(nodes, shape, type, fwd, hash);

src/graph/expression_operators.h

@@ -16,6 +16,11 @@ Expr debug(Expr a, const std::string& message = "");
  */
 Expr checkpoint(Expr a);
 
+/**
+ * Removes the node from the set of root nodes, no-op otherwise
+ */
+Expr removeAsRoot(Expr node);
+
 typedef Expr(ActivationFunction)(Expr);  ///< ActivationFunction has signature Expr(Expr)
 
 /**

src/layers/embedding.cpp

@@ -6,10 +6,8 @@ namespace marian {
 Embedding::Embedding(Ptr<ExpressionGraph> graph, Ptr<Options> options)
     : LayerBase(graph, options), inference_(opt<bool>("inference")) {
   std::string name = opt<std::string>("prefix");
-  int dimVoc       = opt<int>("dimVocab");
-  int dimEmb       = opt<int>("dimEmb");
-  int dimFactorEmb = opt<int>("dimFactorEmb");
-
+  int dimVoc = opt<int>("dimVocab");
+  int dimEmb = opt<int>("dimEmb");
   bool fixed = opt<bool>("fixed", false);
 
   // Embedding layer initialization should depend only on embedding size, hence fanIn=false
@@ -21,6 +19,7 @@ Embedding::Embedding(Ptr<ExpressionGraph> graph, Ptr<Options> options)
     dimVoc = (int)factoredVocab_->factorVocabSize();
     LOG_ONCE(info, "[embedding] Factored embeddings enabled");
     if(opt<std::string>("factorsCombine") == "concat") {
+      int dimFactorEmb = opt<int>("dimFactorEmb", 0);
       ABORT_IF(dimFactorEmb == 0,
                "Embedding: If concatenation is chosen to combine the factor embeddings, a factor "
                "embedding size must be specified.");
@@ -179,8 +178,8 @@ Expr Embedding::applyIndices(const std::vector<WordIndex>& embIdx, const Shape&
       "prefix",         (opt<bool>("tied-embeddings-src") || opt<bool>("tied-embeddings-all")) ? "Wemb"
                                                                                           : prefix_ + "_Wemb",
       "fixed",          embeddingFix_,
-      "dimFactorEmb",   opt<int>("factors-dim-emb"),  // for factored embeddings
-      "factorsCombine", opt<std::string>("factors-combine"),  // for factored embeddings
+      "dimFactorEmb",   opt<int>("factors-dim-emb", 0),  // for factored embeddings
+      "factorsCombine", opt<std::string>("factors-combine", ""),  // for factored embeddings
       "vocab",     opt<std::vector<std::string>>("vocabs")[batchIndex_]);  // for factored embeddings
   // clang-format on
   if(options_->hasAndNotEmpty("embedding-vectors")) {

src/layers/guided_alignment.h

@@ -26,7 +26,8 @@ guidedAlignmentToSparse(Ptr<data::CorpusBatch> batch) {
 
   std::sort(byIndex.begin(), byIndex.end(), [](const BiPoint& a, const BiPoint& b) { return std::get<0>(a) < std::get<0>(b); });
   std::vector<IndexType> indices; std::vector<float> valuesFwd; 
-  indices.reserve(byIndex.size()); valuesFwd.reserve(byIndex.size()); 
+  indices.reserve(byIndex.size()); 
+  valuesFwd.reserve(byIndex.size()); 
   for(auto& p : byIndex) {
     indices.push_back((IndexType)std::get<0>(p));
     valuesFwd.push_back(std::get<1>(p));
@@ -40,28 +41,33 @@ static inline RationalLoss guidedAlignmentCost(Ptr<ExpressionGraph> graph,
                                                Ptr<Options> options,
                                                Expr attention) { // [beam depth=1, max src length, batch size, tgt length]
   std::string guidedLossType = options->get<std::string>("guided-alignment-cost");  // @TODO: change "cost" to "loss"
-
+  // @TODO: It is ugly to check the multi-loss type here, but doing this right requires
+  // a substantial rewrite of the multi-loss architecture, which is planned anyways.
+  std::string multiLossType = options->get<std::string>("multi-loss-type", "sum");
+  
   // We dropped support for other losses which are not possible to implement with sparse labels.
   // They were most likely not used anyway.
   ABORT_IF(guidedLossType != "ce", "Only alignment loss type 'ce' is supported");
 
   float guidedLossWeight = options->get<float>("guided-alignment-weight");
-
-  auto [indices, values]  = guidedAlignmentToSparse(batch);
-  auto alignmentIndices   = graph->indices(indices);
-  auto alignmentValues    = graph->constant({(int)values.size()}, inits::fromVector(values));
-  auto attentionAtAligned = cols(flatten(attention), alignmentIndices);
-
-  float epsilon           = 1e-6f;
-  Expr alignmentLoss      = -sum(cast(alignmentValues * log(attentionAtAligned + epsilon), Type::float32));
-  size_t numLabels        = alignmentIndices->shape().elements();
-
+  const auto& [indices, values] = guidedAlignmentToSparse(batch);
+  
+  Expr alignmentLoss;
+  size_t numLabels = indices.size(); // can be zero
+  if(indices.empty()) {
+    removeAsRoot(stopGradient(attention)); // unused, hence make sure we don't polute the backwards operations
+    alignmentLoss = graph->zeros({1});
+    numLabels = multiLossType == "sum" ? 0 : 1;
+  } else {
+    float epsilon           = 1e-6f;
+    auto alignmentIndices   = graph->indices(indices);
+    auto alignmentValues    = graph->constant({(int)values.size()}, inits::fromVector(values));
+    auto attentionAtAligned = cols(flatten(attention), alignmentIndices);
+    alignmentLoss           = -sum(cast(alignmentValues * log(attentionAtAligned + epsilon), Type::float32));
+  }
   // Create label node, also weigh by scalar so labels and cost are in the same domain.
   // Fractional label counts are OK. But only if combined as "sum".
-  // @TODO: It is ugly to check the multi-loss type here, but doing this right requires
-  // a substantial rewrite of the multi-loss architecture, which is planned anyways.
-  std::string multiLossType = options->get<std::string>("multi-loss-type", "sum");
-  if (multiLossType == "sum")         // sum of sums
+  if (multiLossType == "sum") // sum of sums
     return RationalLoss(guidedLossWeight * alignmentLoss, guidedLossWeight * numLabels);
   else
     return RationalLoss(guidedLossWeight * alignmentLoss, (float)numLabels);

src/layers/lsh.cpp

@@ -1,6 +1,12 @@
-#include "layers/lsh.h"
-#include "tensors/tensor_operators.h"
+#include "common/timer.h"
 #include "common/utils.h"
+#include "layers/lsh.h"
+#include "layers/lsh_impl.h"
+#include "tensors/tensor_operators.h"
+
+#if _MSC_VER
+#include "3rd_party/faiss/Index.h"
+#endif
 
 #include "3rd_party/faiss/utils/hamming.h"
 
@@ -8,10 +14,6 @@
 #include "3rd_party/faiss/VectorTransform.h"
 #endif
 
-#include "common/timer.h"
-
-#include "layers/lsh_impl.h"
-
 namespace marian {
 namespace lsh {
 
@@ -116,7 +118,7 @@ Expr searchEncoded(Expr encodedQuery, Expr encodedWeights, int dimK, int firstNR
 
   int currBeamSize = encodedQuery->shape()[0];
   int batchSize    = encodedQuery->shape()[2];
-
+  
   auto search = [=](Expr out, const std::vector<Expr>& inputs) {
     Expr encodedQuery   = inputs[0];
     Expr encodedWeights = inputs[1];
@@ -130,6 +132,32 @@ Expr searchEncoded(Expr encodedQuery, Expr encodedWeights, int dimK, int firstNR
 
     ABORT_IF(dimK > wRows, "k is larger than number of candidate values?"); // @TODO: use min(k, wRows) silently?
 
+#if _MSC_VER // unfortunately MSVC is horrible at loop unrolling, so we fall back to the old code (hrmph!) @TODO: figure this out one day
+    int qRows = encodedQuery->shape().elements() / bytesPerVector;
+
+    uint8_t* qCodes = encodedQuery->val()->data<uint8_t>();
+    uint8_t* wCodes = encodedWeights->val()->data<uint8_t>();
+
+    // use actual faiss code for performing the hamming search. 
+    std::vector<int> distances(qRows * dimK);
+    std::vector<faiss::Index::idx_t> ids(qRows * dimK);
+    faiss::int_maxheap_array_t res = {(size_t)qRows, (size_t)dimK, ids.data(), distances.data()};
+    faiss::hammings_knn_hc(&res, qCodes, wCodes, (size_t)wRows, (size_t)bytesPerVector, 0);
+
+    // Copy int64_t indices to Marian index type and sort by increasing index value per hypothesis.
+    // The sorting is required as we later do a binary search on those values for reverse look-up.
+    uint32_t* outData = out->val()->data<uint32_t>();
+    
+    int numHypos = out->shape().elements() / dimK;
+    for (size_t hypoIdx = 0; hypoIdx < numHypos; ++hypoIdx) {
+      size_t startIdx = dimK * hypoIdx;
+      size_t endIdx = startIdx + dimK;
+      for(size_t i = startIdx; i < endIdx; ++i)
+        outData[i] = (uint32_t)ids[i];
+      if(!noSort)
+        std::sort(outData + startIdx, outData + endIdx);
+    }
+#else // this is using the new code for search, other parts of the code, like conversion are fine.
     IndexType* outData = out->val()->data<IndexType>();
     auto gather = [outData, dimK](IndexType rowId, IndexType k, IndexType kthColId, DistType /*dist*/) { 
       outData[rowId * dimK + k] = kthColId; 
@@ -144,6 +172,7 @@ Expr searchEncoded(Expr encodedQuery, Expr encodedWeights, int dimK, int firstNR
     params.bytesPerVector = bytesPerVector;
 
     hammingTopK(params, gather);
+#endif
   };
 
   Shape kShape({currBeamSize, batchSize, dimK});

src/training/scheduler.h

@@ -494,12 +494,17 @@ public:
       state_->wordsDisp    = 0;
     }
 
-    if(options_->get<bool>("valid-reset-stalled")) {
+    if(options_->get<bool>("valid-reset-stalled") || options_->get<bool>("valid-reset-all")) {
       state_->stalled      = 0;
       state_->maxStalled   = 0;
       for(const auto& validator : validators_) {
-        if(state_->validators[validator->type()])
+        if(state_->validators[validator->type()]) {
+          // reset the number of stalled validations, e.g. when the validation set is the same
           state_->validators[validator->type()]["stalled"] = 0;
+          // reset last best results as well, e.g. when the validation set changes
+          if(options_->get<bool>("valid-reset-all"))
+            state_->validators[validator->type()]["last-best"] = validator->initScore();
+        }
       }
     }
 
@@ -512,10 +517,10 @@ public:
     if(mpi_->isMainProcess())
       if(filesystem::exists(nameYaml))
         yamlStr = io::InputFileStream(nameYaml).readToString();
-    
+
     if(mpi_)
       mpi_->bCast(yamlStr);
-    
+
     loadFromString(yamlStr);
   }