Merge branch 'lstm' into 'dev'

Fixed issues for LSTM

See merge request !267

include/aidge/operator/Stack.hpp

@@ -50,7 +50,7 @@ class StackOp
   public:
     static const std::string s_type;
 
-    StackOp(std::uint32_t maxElements);
+    StackOp(std::uint32_t maxElements = 0);
 
     /**
      * @brief Copy-constructor. Copy the operator attributes and its output
@@ -71,6 +71,7 @@ class StackOp
 
     std::set<std::string> getAvailableBackends() const override;
 
+    bool dimsForwarded() const override final;
     bool forwardDims(bool allowDataDependency = false) override final;
     void forward() override;
 
@@ -87,14 +88,14 @@ class StackOp
     }
 
     static const std::vector<std::string> getInputsName() {
-        return {"data_input"};
+        return {"data_input", "max_elements"};
     }
     static const std::vector<std::string> getOutputsName() {
         return {"data_output"};
     }
 };
 
-std::shared_ptr<Node> stack(std::uint32_t maxElements,
+std::shared_ptr<Node> Stack(std::uint32_t maxElements = 0,
                             const std::string &name = "");
 } // namespace Aidge
 

python_binding/operator/pybind_Stack.cpp

@@ -29,8 +29,8 @@ void init_Stack(py::module &m) {
         .def_readonly_static("Type", &StackOp::s_type);
 
     m.def("Stack",
-          &stack,
-          py::arg("max_elements"),
+          &Stack,
+          py::arg("max_elements") = 0,
           py::arg("name") = "",
           R"mydelimiter(
         Initialize a node containing a Stack operator.

src/operator/Shape.cpp

@@ -20,14 +20,8 @@
 #include "aidge/utils/Types.h"
 
 void Aidge::Shape_OpImpl::forward() {
-    const Shape_Op& op = dynamic_cast<const Shape_Op&>(mOp);
-    const auto start = op.start();
-    const auto end = op.end();
-
-    op.getOutput(0)->getImpl()->copyCast(std::next(op.getInput(0)->dims().data(),
-                                                   start),
-                                         DataType::UInt64,
-                                         end - start + 1);
+    // Do nothing...
+    // Output is already valid after forwardDims()
 }
 
 ///////////////////////////////////////////////
@@ -75,6 +69,11 @@ bool Aidge::Shape_Op::forwardDims(bool /*allowDataDependency*/) {
         AIDGE_ASSERT(roi> 1, "Invalid ROI for Shape");
 
         mOutputs[0]->resize({roi});
+        // Ensure the output of this operator is valid after forwardDims():
+        mOutputs[0]->getImpl()->copyCast(std::next(getInput(0)->dims().data(),
+                                                    start),
+                                            DataType::UInt64,
+                                            end - start + 1);
         return true;
     }
 

src/operator/Stack.cpp

@@ -26,7 +26,7 @@ namespace Aidge {
 // inputSize
 Elts_t StackProdConso::getRequiredMemory(
     const Aidge::IOIndex_t inputIdx,
-    const std::vector<DimSize_t> &inputsSize) const {
+    const std::vector<DimSize_t> &/*inputsSize*/) const {
     assert(mOp.getRawInput(inputIdx) && "requires valid input");
 
     const StackOp &op = dynamic_cast<const StackOp &>(mOp);
@@ -62,15 +62,10 @@ void StackOpImpl::forward() {
 }
 
 StackOp::StackOp(std::uint32_t maxElements)
-    : OperatorTensor(s_type, {InputCategory::Data}, 1),
+    : OperatorTensor(s_type, {InputCategory::Data, InputCategory::OptionalData}, 1),
       mAttributes(std::make_shared<Attributes_>(
           attr<StackAttr::MaxElements>(maxElements),
           attr<StackAttr::ForwardStep>(0))) {
-    if (maxElements == 0) {
-        AIDGE_THROW_OR_ABORT(
-            std::invalid_argument,
-            "StackOp creation failed: maxElements must be greater than 0.");
-    }
     mImpl = std::make_shared<StackOpImpl>(*this);
 }
 
@@ -87,8 +82,33 @@ std::shared_ptr<Aidge::Operator> Aidge::StackOp::clone() const {
     return std::make_shared<StackOp>(*this);
 }
 
-bool Aidge::StackOp::forwardDims(bool /*allowDataDependency*/) {
+bool Aidge::StackOp::dimsForwarded() const {
+    if ((getInput(1) && !getInput(1)->undefined()))
+    {
+        // output dims are data dependent
+        return false;
+    }
+
+    return OperatorTensor::dimsForwarded();
+}
+
+bool Aidge::StackOp::forwardDims(bool allowDataDependency) {
     if (inputsAssociated()) {
+        // Copy optional input #1 first dimension, if present, to attribute MaxElements
+        if (getInput(1)) {
+            if (!allowDataDependency) {
+                Log::warn("StackOp: unable to forwardDims() because output dims are data dependent on input#1");
+                return false;
+            }
+
+            std::shared_ptr<Tensor> fallback;
+            const auto& maxElements = getInput(1)->refCastFrom(fallback, NativeType<std::uint32_t>::type, "cpu");
+            AIDGE_ASSERT(maxElements.size() > 0, "Input#1 size should be > 0");
+            this->maxElements() = static_cast<std::uint32_t*>(maxElements.getImpl()->hostPtr())[0];
+        }
+
+        AIDGE_ASSERT(this->maxElements() > 0, "Input#1 first element or MaxElements attribute should be > 0");
+
         auto inputDims = getInput(0)->dims();
         inputDims.insert(inputDims.begin(), maxElements());
         getOutput(0)->resize(inputDims);
@@ -116,7 +136,7 @@ void StackOp::forward() {
     ++forwardStep();
 }
 
-std::shared_ptr<Node> stack(std::uint32_t maxElements,
+std::shared_ptr<Node> Stack(std::uint32_t maxElements,
                             const std::string &name) {
     return std::make_shared<Node>(std::make_shared<StackOp>(maxElements),
                                   name);

unit_tests/operator/Test_StackImpl.cpp

@@ -56,9 +56,6 @@ TEST_CASE("[core/operator] Stack(forward)", "[Stack]") {
             REQUIRE(op2.maxElements() == maxElements);
             REQUIRE(op2.forwardStep() == 0);
         }
-
-        // Invalid arguments
-        REQUIRE_THROWS_AS(StackOp(0), std::invalid_argument);
     }
 
     SECTION("forwardDims") {
@@ -111,7 +108,7 @@ TEST_CASE("[core/operator] Stack(forward)", "[Stack]") {
             tensors[i]->getImpl()->setRawPtr(arrays[i], nbElems);
         }
 
-        auto myStack = stack(numTensors);
+        auto myStack = Stack(numTensors);
         myStack->getOperator()->setBackend("cpu");
         myStack->getOperator()->setDataType(DataType::Float32);
         auto op =