Merge branch 'tiling' into 'main'

Update fuseBatchNorm to include ConvDepthWise and MetaOperators

See merge request !62

include/aidge/graph/GraphView.hpp

@@ -96,7 +96,7 @@ public:
      * specified location.
      * @param path
      */
-    void save(std::string path, bool verbose = false) const;
+    void save(std::string path, bool verbose = false, bool showProducers = true) const;
 
     inline bool inView(NodePtr nodePtr) const {
         return mNodes.find(nodePtr) != mNodes.end();

include/aidge/operator/Erf.hpp

@@ -51,12 +51,9 @@ public:
         return std::make_shared<Erf_Op>(*this);
     }
 
-    void setBackend(const std::string& name) override {
+    void setBackend(const std::string& name, DeviceIdx_t device = 0) override {
         mImpl = Registrar<Erf_Op>::create(name)(*this);
-        mOutputs[0]->setBackend(name);
-
-        // FIXME: temporary workaround
-        getInput(0)->setBackend(name);
+        mOutputs[0]->setBackend(name, device);
     }
 
     static const std::vector<std::string> getInputsName(){

include/aidge/operator/Gather.hpp

@@ -40,7 +40,7 @@ public:
 
     Gather_Op() = delete;
 
-    
+
     using Attributes_ = StaticAttributes<GatherAttr, int>;
     template <GatherAttr e> using attr = typename Attributes_::template attr<e>;
     Gather_Op(int axis)
@@ -70,13 +70,9 @@ public:
 
     void computeOutputDims() override final;
 
-    void setBackend(const std::string& name) override {
+    void setBackend(const std::string& name, DeviceIdx_t device = 0) override {
         mImpl = Registrar<Gather_Op>::create(name)(*this);
-        mOutputs[0]->setBackend(name);
-
-        // FIXME: temporary workaround
-        getInput(0)->setBackend(name);
-        getInput(1)->setBackend(name);
+        mOutputs[0]->setBackend(name, device);
     }
 
     static const std::vector<std::string> getInputsName(){

include/aidge/operator/MetaOperator.hpp

@@ -46,11 +46,11 @@ public:
         return std::make_shared<MetaOperator_Op>(*this);
     }
 
-    const std::shared_ptr<GraphView>& getMicroGraph() const {
+    inline const std::shared_ptr<GraphView>& getMicroGraph() const noexcept {
         return mGraph;
     }
 
-    const std::shared_ptr<SequentialScheduler>& getMicroGraphScheduler() const {
+    inline const std::shared_ptr<SequentialScheduler>& getMicroGraphScheduler() const noexcept {
         return mScheduler;
     }
 

include/aidge/operator/Operator.hpp

@@ -73,12 +73,18 @@ public:
 public:
     virtual std::shared_ptr<Operator> clone() const = 0;
 
+    /**
+     * @brief Set the specified input with a shallow copy.
+     * @param inputIdx Index of the input to set.
+     * @param data Data to copy.
+     */
     virtual void associateInput(const IOIndex_t inputIdx, const std::shared_ptr<Data>& data) = 0;
 
     /**
      * @brief Set the specified input by performing a deep copy of the given data.
      * The pointer itself is not changed, thus keeping the current connections.
      * @param inputIdx Index of the input to set.
+     * @param data Data to copy.
      */
     virtual void setInput(const IOIndex_t inputIdx, const std::shared_ptr<Data>& data) = 0;
     virtual void setInput(const IOIndex_t inputIdx, std::shared_ptr<Data>&& data) = 0;

include/aidge/operator/ReduceMean.hpp

@@ -89,7 +89,7 @@ class ReduceMean_Op : public OperatorTensor,
                 }
                 else
                     outDims.push_back(getInput(0)->dims()[d]);
-            }        
+            }
             if(outDims.size()>0)
                 mOutputs[0]->resize(outDims);
             else
@@ -97,12 +97,9 @@ class ReduceMean_Op : public OperatorTensor,
         }
     }
 
-    void setBackend(const std::string &name) override {
+    void setBackend(const std::string &name, DeviceIdx_t device = 0) override {
         mImpl = Registrar<ReduceMean_Op<DIM>>::create(name)(*this);
-        mOutputs[0]->setBackend(name);
-
-        // FIXME: temporary workaround
-        getInput(0)->setBackend(name);
+        mOutputs[0]->setBackend(name, device);
     }
 
     static const std::vector<std::string> getInputsName(){

include/aidge/operator/Reshape.hpp

@@ -66,12 +66,9 @@ public:
 
     void computeOutputDims() override final;
 
-    void setBackend(const std::string& name) override {
+    void setBackend(const std::string& name, DeviceIdx_t device = 0) override {
         mImpl = Registrar<Reshape_Op>::create(name)(*this);
-        mOutputs[0]->setBackend(name);
-
-        // FIXME: temporary workaround
-        getInput(0)->setBackend(name);
+        mOutputs[0]->setBackend(name, device);
     }
 
     static const std::vector<std::string> getInputsName(){

include/aidge/operator/Transpose.hpp

@@ -79,12 +79,9 @@ class Transpose_Op : public OperatorTensor,
         }
     }
 
-    void setBackend(const std::string &name) override {
+    void setBackend(const std::string &name, DeviceIdx_t device = 0) override {
         mImpl = Registrar<Transpose_Op<DIM>>::create(name)(*this);
-        mOutputs[0]->setBackend(name);
-
-        // FIXME: temporary workaround
-        getInput(0)->setBackend(name);
+        mOutputs[0]->setBackend(name, device);
     }
 
     static const std::vector<std::string> getInputsName(){

python_binding/graph/pybind_GraphView.cpp

@@ -23,7 +23,7 @@ namespace Aidge {
 void init_GraphView(py::module& m) {
     py::class_<GraphView, std::shared_ptr<GraphView>>(m, "GraphView")
           .def(py::init<>())
-          .def("save", &GraphView::save, py::arg("path"), py::arg("verbose") = false,
+          .def("save", &GraphView::save, py::arg("path"), py::arg("verbose") = false, py::arg("show_producers") = true,
           R"mydelimiter(
           Save the GraphView as a Mermaid graph in a .md file at the specified location.
 
@@ -97,7 +97,7 @@ void init_GraphView(py::module& m) {
           .def("get_nodes", &GraphView::getNodes)
           .def("get_node", &GraphView::getNode, py::arg("node_name"))
           .def("forward_dims", &GraphView::forwardDims)
-          .def("compile", &GraphView::compile, py::arg("backend"), py::arg("datatype"))
+          .def("compile", &GraphView::compile, py::arg("backend"), py::arg("datatype"), py::arg("device") = 0)
           .def("__call__", &GraphView::operator(), py::arg("connectors"))
           .def("set_datatype", &GraphView::setDataType, py::arg("datatype"))
           .def("set_backend", &GraphView::setBackend, py::arg("backend"), py::arg("device") = 0)

src/graph/GraphView.cpp

@@ -55,7 +55,7 @@ std::string Aidge::GraphView::name() const { return mName; }
 void Aidge::GraphView::setName(const std::string &name) { mName = name; }
 
 
-void Aidge::GraphView::save(std::string path, bool verbose) const {
+void Aidge::GraphView::save(std::string path, bool verbose, bool showProducers) const {
     FILE *fp = std::fopen((path + ".mmd").c_str(), "w");
     std::fprintf(fp,
                 "%%%%{init: {'flowchart': { 'curve': 'monotoneY'}, "
@@ -68,7 +68,7 @@ void Aidge::GraphView::save(std::string path, bool verbose) const {
     for (const std::shared_ptr<Node> &node_ptr : mNodes) {
         const std::string currentType = node_ptr->type();
         if (typeCounter.find(currentType) == typeCounter.end())
-        typeCounter[currentType] = 0;
+            typeCounter[currentType] = 0;
         ++typeCounter[currentType];
 
         std::string givenName =
@@ -83,13 +83,18 @@ void Aidge::GraphView::save(std::string path, bool verbose) const {
                       givenName.c_str());
         }
         else {
-          std::fprintf(fp, "%s(%s)\n", namePtrTable[node_ptr].c_str(),
-                      givenName.c_str());
+            if ((currentType != "Producer") || showProducers) {
+                std::fprintf(fp, "%s(%s)\n", namePtrTable[node_ptr].c_str(),
+                            givenName.c_str());
+            }
         }
     }
 
     // Write every link
     for (const std::shared_ptr<Node> &node_ptr : mNodes) {
+      if ((node_ptr -> type() == "Producer") && !showProducers) {
+        continue;
+      }
       IOIndex_t outputIdx = 0;
       for (auto childs : node_ptr->getOrderedChildren()) {
         for (auto child : childs) {

src/graphRegex/GraphRegex.cpp

 #include "aidge/graphRegex/GraphRegex.hpp"
-using namespace Aidge; 
+using namespace Aidge;
 
 
 void GraphRegex::setKeyFromGraph(std::shared_ptr<GraphView> ref){
@@ -27,7 +27,7 @@ void GraphRegex::setKeyFromGraph(std::shared_ptr<GraphView> ref){
 
 
 // void GraphRegex::addQuery(const std::string query){
-//     //TODO one query only but the same string is a same query but 
+//     //TODO one query only but the same string is a same query but
 //     //2 different string it's maybe the same query , we need to check the AST
 //     mQueryRecipe[query] = nullptr;
 // }
@@ -52,7 +52,7 @@ void GraphRegex::_generateCombinationsStart(const std::set<NodePtr>& elements, s
     }
 }
 
-
+// factorial(n) tree searched optimized with a stopping condition
 void GraphRegex::_findLargestCompatibleSet(
     const std::vector<std::shared_ptr<MatchSolution>>& solutions,
     std::set<std::shared_ptr<MatchSolution>>& currentSet,
@@ -75,6 +75,10 @@ void GraphRegex::_findLargestCompatibleSet(
             currentSet.insert(solutions[i]);
             _findLargestCompatibleSet(solutions, currentSet, largestSet, i + 1);
             currentSet.erase(solutions[i]);
+            // cut the size of the graph of possibilities
+            if ((currentSet.size() + solutions.size() - currentIndex) <= largestSet.size()) {
+                return;
+            }
         }
     }
 }
@@ -101,14 +105,14 @@ std::set<std::shared_ptr<MatchSolution>> GraphRegex::match(std::shared_ptr<Graph
         std::shared_ptr<GraphFsmInterpreter>  fsmGenerator = std::make_shared<GraphFsmInterpreter>(query,mAllTest);
         std::shared_ptr<FsmGraph> fsm = fsmGenerator->interpret();
 
-        // generate all the start possibility 
+        // generate all the start possibility
         std::size_t nb_startSt =  fsm->getNbStart();
         std::set<std::vector<NodePtr>> combinations;
         std::vector<NodePtr> current;
         _generateCombinationsStart(ref->getNodes(), nb_startSt, 0, current, combinations);
-        
 
-        // all start 
+
+        // all start
         for (const auto& combination : combinations) {
             std::vector<std::shared_ptr<MatchSolution>> solution = fsm->test(combination);
             solutions.insert(solutions.end(), solution.begin(), solution.end());
@@ -133,7 +137,7 @@ void GraphRegex::setNodeKey(const std::string key, const std::string conditional
 
 
 void GraphRegex::setNodeKey(const std::string key,std::function<bool(NodePtr)> f){
-    //we can applied to all key but it's not efficient 
+    //we can applied to all key but it's not efficient
     if(mAllLambda.find(key) != mAllLambda.end()){
         throw std::runtime_error(key + " is define");
     }
@@ -142,7 +146,7 @@ void GraphRegex::setNodeKey(const std::string key,std::function<bool(NodePtr)> f
 }
 
 void GraphRegex::_majConditionalInterpreterLambda(){
-       
+
     for (const auto& test : mAllTest) {
         for (const auto& pair : mAllLambda) {
             const std::string& key = pair.first;
@@ -151,7 +155,7 @@ void GraphRegex::_majConditionalInterpreterLambda(){
             if(!test->isLambdaRegister(key)){
                 test->insertLambda(key,lambda);
             }
-            
+
         }
     }
 }

src/recipies/FuseBatchNorm.cpp

@@ -8,30 +8,66 @@
  * SPDX-License-Identifier: EPL-2.0
  *
  ********************************************************************************/
-#include <set>
 #include <cassert>
 #include <memory>
+#include <set>
 #include <string>
 
-#include "aidge/operator/FC.hpp"
+#include "aidge/data/Tensor.hpp"
+#include "aidge/graph/GraphView.hpp"
+#include "aidge/graph/Node.hpp"
 #include "aidge/operator/BatchNorm.hpp"
 #include "aidge/operator/Conv.hpp"
+#include "aidge/operator/ConvDepthWise.hpp"
+#include "aidge/operator/FC.hpp"
+#include "aidge/operator/MetaOperator.hpp"
 #include "aidge/recipies/Recipies.hpp"
-#include "aidge/graph/GraphView.hpp"
-#include "aidge/graph/Node.hpp"
-#include "aidge/operator/Producer.hpp"
-#include "aidge/operator/GenericOperator.hpp"
-
+#include "aidge/utils/ErrorHandling.hpp"
+#include "aidge/utils/Types.h"
 
-//Graph Regex
+// Graph Regex
 #include "aidge/graphRegex/GraphRegex.hpp"
 
-void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::Node> convNode, std::shared_ptr<Aidge::Node> batchnormNode) {
+void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::Node> convNode,
+                          std::shared_ptr<Aidge::Node> batchnormNode) {
+    // Case: convNode is a MetaOperator ending with a Convolution
+    // eg. PaddedConv
+    if (!(convNode -> getOperator() -> isAtomic())) {
+        const std::shared_ptr<MetaOperator_Op> metaNode = std::static_pointer_cast<MetaOperator_Op>(convNode -> getOperator());
+        const std::shared_ptr<GraphView>  metanodeGraph = metaNode -> getMicroGraph();
+        const std::vector<std::pair<std::shared_ptr<Node>, IOIndex_t>> outputNodes = metanodeGraph -> getOrderedOutputs();
+        if (outputNodes.size() != 1) {
+            AIDGE_THROW_OR_ABORT(std::runtime_error, "Bad MetaOperator argument for fuseBatchNorm recipie.");
+        }
+        convNode = outputNodes[0].first;
+    }
+
+    AIDGE_ASSERT(((convNode->type() == Conv_Op<2>::Type) || (convNode->type() == ConvDepthWise_Op<2>::Type)), "Wrong type");
+    AIDGE_ASSERT(batchnormNode->type() == BatchNorm_Op<2>::Type, "Wrong type for batchnorm node.");
 
     // TODO: Find a way to remove the template
     // A feature map with 2 dimensions is assumed
-    const std::shared_ptr<BatchNorm_Op<2>> batchOp = std::static_pointer_cast<BatchNorm_Op<2>>(batchnormNode->getOperator());
-    const std::shared_ptr<Conv_Op<2>> convOp = std::static_pointer_cast<Conv_Op<2>>(convNode->getOperator());
+    const std::shared_ptr<BatchNorm_Op<2>> batchOp =
+            std::static_pointer_cast<BatchNorm_Op<2>>(batchnormNode->getOperator());
+
+    DimSize_t convNbOutChannels;
+    DimSize_t channelsSize;
+    std::array<DimSize_t, 2> kernelDims;
+    std::shared_ptr<OperatorTensor> convOp = std::static_pointer_cast<OperatorTensor>(convNode->getOperator());
+    if (convNode->type() == Conv_Op<2>::Type) {
+        const std::shared_ptr<Conv_Op<2>> convOpPtr =
+            std::static_pointer_cast<Conv_Op<2>>(convNode->getOperator());
+        convNbOutChannels = convOpPtr->getAttr<DimSize_t>("OutChannels");
+        channelsSize = convOpPtr->getAttr<DimSize_t>("InChannels");
+        kernelDims = convOpPtr->getAttr<std::array<DimSize_t, 2>>("KernelDims");
+    }
+    else if (convNode->type() == ConvDepthWise_Op<2>::Type) {
+        const std::shared_ptr<ConvDepthWise_Op<2>> convOpPtr =
+            std::static_pointer_cast<ConvDepthWise_Op<2>>(convNode->getOperator());
+        convNbOutChannels = convOpPtr->getAttr<DimSize_t>("Channels");
+        channelsSize = 1;
+        kernelDims = convOpPtr->getAttr<std::array<DimSize_t, 2>>("KernelDims");
+    }
 
     std::shared_ptr<Tensor> scaleBuf, shiftBuf, b_meanBuf, b_varBuf;
     const Tensor& scale = batchOp->getInput(1)->refCastFrom(scaleBuf, DataType::Float32, "cpu");
@@ -39,20 +75,12 @@ void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::Node> convNode, std::shared_ptr
     const Tensor& b_mean = batchOp->getInput(3)->refCastFrom(b_meanBuf, DataType::Float32, "cpu");
     const Tensor& b_var = batchOp->getInput(4)->refCastFrom(b_varBuf, DataType::Float32, "cpu");
 
-    const float epsilon = batchOp -> getAttr<float>("Epsilon");
-    const DimSize_t convNbOutChannels = convOp -> getAttr<DimSize_t>("OutChannels");
-    const DimSize_t channelsSize = convOp -> getAttr<DimSize_t>("InChannels");
-    const std::array<DimSize_t, 2> kernelDims = convOp -> getAttr<std::array<DimSize_t, 2>>("KernelDims");
+    const float epsilon = batchOp->getAttr<float>("Epsilon");
 
 
-    assert(scale.size()  == convNbOutChannels);
-    assert(shift.size()  == convNbOutChannels);
-    assert(b_mean.size() == convNbOutChannels);
-    assert(b_var.size()  == convNbOutChannels);
     assert(epsilon > 0.0);
     // TODO : no no_bias attribute ?
 
-
     float meanVariance = 0.0;
     unsigned int count = 0;
 
@@ -60,8 +88,7 @@ void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::Node> convNode, std::shared_ptr
         if (b_var.get<float>(outChId) > 1.0e-12) {
             meanVariance += b_var.get<float>(outChId);
             ++count;
-        }
-        else {
+        } else {
             printf("Zero-variance: %s [%lu]\n", convNode->name().c_str(), outChId);
         }
     }
@@ -86,8 +113,8 @@ void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::Node> convNode, std::shared_ptr
         // Weights adjustments
         for (std::size_t channel = 0; channel < channelsSize; ++channel) {
             // TODO : Suppose kerneldims = 2
-            for(std::size_t k0 = 0; k0 < kernelDims[0]; ++ k0){
-                for(std::size_t k1 = 0; k1 < kernelDims[1]; ++ k1){
+            for (std::size_t k0 = 0; k0 < kernelDims[0]; ++k0) {
+                for (std::size_t k1 = 0; k1 < kernelDims[1]; ++k1) {
                     std::vector<DimSize_t> currentIdx = {outChId, channel, k0, k1};
                     float weightValue = weight.get<float>(currentIdx);
                     weight.set<float>(currentIdx, weightValue*factor); // Update check it update Conv weights
@@ -119,24 +146,37 @@ void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::Node> convNode, std::shared_ptr
 }
 
 void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::MatchSolution> solution) {
-
     assert(solution->at("BatchNorm").size() == 1 && "Wrong number of nodes BatchNorm to replace\n");
     assert(solution->at("OP").size() == 1 && "Wrong number of nodes OP to replace\n");
 
     for (const auto& op : solution->at("OP")) {
-        for (const auto& batchNorm : solution->at("BatchNorm")) {
-            fuseBatchNorm(op,batchNorm);
+        if (op->getOperator()->isAtomic()) {
+            for (const auto& batchNorm : solution->at("BatchNorm")) {
+                fuseBatchNorm(op, batchNorm);
+            }
+        } else {  // op is a MetaOperator
+            auto metaOp = std::dynamic_pointer_cast<MetaOperator_Op>(op->getOperator());
+            if ((metaOp->getMicroGraph()->getOrderedOutputs().size() == 1) &&
+                ((metaOp->getMicroGraph()->getOrderedOutputs()[0].first->type() ==
+                  Conv_Op<2>::Type) ||
+                 (metaOp->getMicroGraph()->getOrderedOutputs()[0].first->type() ==
+                  ConvDepthWise_Op<2>::Type))) {
+                for (const auto& batchNorm : solution->at("BatchNorm")) {
+                    fuseBatchNorm(op, batchNorm);
+                }
+            }
         }
     }
-
 }
 
 void Aidge::fuseBatchNorm(std::shared_ptr<Aidge::GraphView> graphView) {
-
-
     std::shared_ptr<GraphRegex> regex = std::make_shared<GraphRegex>();
-    regex->setNodeKey("BatchNorm","getType($) =='BatchNorm'");
-    regex->setNodeKey("OP","getType($) =='Conv'");//  || getType($) =='FC' ");
+    regex->setNodeKey("BatchNorm", "getType($) =='BatchNorm'");
+    printf("\n============================\nSearching for solutions\n==============================\n");
+    regex->setNodeKey(
+            "OP",
+            "getType($) =='Conv' || getType($) =='ConvDepthWise' || getType($) =='PaddedConv' || getType($) =='PaddedConvDepthWise'");
+            //  || getType($) =='FC' ");
 
     regex->addQuery("OP -> BatchNorm");