From 2ac231d8a2e633b4439607a6065ebe919c264d29 Mon Sep 17 00:00:00 2001
From: NAUD Maxence <maxence.naud@cea.fr>
Date: Thu, 19 Oct 2023 08:35:44 +0000
Subject: [PATCH] [Upd] Add Operator to support an undefined number of inputs.
Remove templates
---
include/aidge/aidge.hpp | 2 +-
include/aidge/operator/Add.hpp | 66 ++++++++++++-------------
python_binding/operator/pybind_Add.cpp | 12 ++---
unit_tests/recipies/Test_FuseMulAdd.cpp | 4 +-
4 files changed, 41 insertions(+), 43 deletions(-)
diff --git a/include/aidge/aidge.hpp b/include/aidge/aidge.hpp
index 16fa9967c..8a1b50a0e 100644
--- a/include/aidge/aidge.hpp
+++ b/include/aidge/aidge.hpp
@@ -33,7 +33,7 @@
#include "aidge/operator/Add.hpp"
#include "aidge/operator/AvgPooling.hpp"
#include "aidge/operator/BatchNorm.hpp"
-#include "aidge/operator/Concat.hpp"
+// #include "aidge/operator/Concat.hpp"
#include "aidge/operator/Conv.hpp"
#include "aidge/operator/ConvDepthWise.hpp"
#include "aidge/operator/FC.hpp"
diff --git a/include/aidge/operator/Add.hpp b/include/aidge/operator/Add.hpp
index 65c7e8ce0..ceb058dbd 100644
--- a/include/aidge/operator/Add.hpp
+++ b/include/aidge/operator/Add.hpp
@@ -16,7 +16,7 @@
#include <vector>
#include <cmath>
#include <memory>
-#include <array>
+#include <vector>
#include "aidge/utils/Registrar.hpp"
#include "aidge/operator/Operator.hpp"
@@ -26,24 +26,23 @@
namespace Aidge {
-template <std::size_t NUM>
class Add_Op : public Operator,
- public Registrable<Add_Op<NUM>, std::string, std::unique_ptr<OperatorImpl>(const Add_Op<NUM>&)> {
-public:
+ public Registrable<Add_Op, std::string, std::unique_ptr<OperatorImpl>(const Add_Op&)> {
+private:
// FIXME: change accessibility
- std::array<std::shared_ptr<Tensor>, NUM> mInputs;
+ std::vector<std::shared_ptr<Tensor>> mInputs;
const std::shared_ptr<Tensor> mOutput = std::make_shared<Tensor>();
+ const IOIndex_t mNbInputs;
public:
static constexpr const char* Type = "Add";
- constexpr Add_Op()
- : Operator(Type)
+ Add_Op(const IOIndex_t nbIn)
+ : Operator(Type),
+ mInputs(std::vector<std::shared_ptr<Tensor>>(nbIn, std::make_shared<Tensor>())),
+ mNbInputs(nbIn)
{
- assert(NUM > 0 && "Add should have at least one input");
- for (std::size_t i = 0; i<NUM; ++i) {
- mInputs[i] = std::make_shared<Tensor>();
- }
+ assert(nbIn > 0 && "Add should have at least one input");
setDatatype(DataType::Float32);
}
@@ -51,17 +50,16 @@ public:
* @brief Copy-constructor. Copy the operator attributes and its output tensor(s), but not its input tensors (the new operator has no input associated).
* @param op Operator to copy.
*/
- Add_Op(const Add_Op<NUM>& op)
+ Add_Op(const Add_Op& op)
: Operator(Type),
+ mInputs(op.mInputs),
+ mNbInputs(op.mNbInputs),
mOutput(std::make_shared<Tensor>(*op.mOutput))
{
// cpy-ctor
- assert(NUM > 0 && "Add should have at least one input");
- for (std::size_t i = 0; i<NUM; ++i) {
- mInputs[i] = std::make_shared<Tensor>();
- }
+ assert(mNbInputs > 0 && "Add should have at least one input");
setDatatype(op.mOutput->dataType());
- mImpl = op.mImpl ? Registrar<Add_Op<NUM>>::create(mOutput->getImpl()->backend())(*this) : nullptr;
+ mImpl = op.mImpl ? Registrar<Add_Op>::create(mOutput->getImpl()->backend())(*this) : nullptr;
}
/**
@@ -82,7 +80,7 @@ public:
// }
void associateInput(const IOIndex_t inputIdx, std::shared_ptr<Data> data) override final {
- assert(static_cast<std::size_t>(inputIdx) < NUM && "wrong inputIdx for Add operator.");
+ assert(static_cast<std::size_t>(inputIdx) < mNbInputs && "wrong inputIdx for Add operator.");
assert(strcmp(data->type(), Tensor::Type)==0 && "input data must be of Tensor type");
mInputs[inputIdx] = std::dynamic_pointer_cast<Tensor>(data);
@@ -92,10 +90,10 @@ public:
if (!mInputs[0]->empty()) {
const auto expectedDims = mInputs[0]->dims();
std::size_t nonEmptyInputTensor = 1;
- for (; nonEmptyInputTensor<NUM && (!mInputs[nonEmptyInputTensor]->empty()); ++nonEmptyInputTensor) {
+ for (; nonEmptyInputTensor < mNbInputs && (!mInputs[nonEmptyInputTensor]->empty()); ++nonEmptyInputTensor) {
assert(expectedDims == mInputs[nonEmptyInputTensor]->dims());
}
- if (nonEmptyInputTensor == NUM) {
+ if (nonEmptyInputTensor == mNbInputs) {
mOutput->resize(expectedDims);
}
}
@@ -103,8 +101,8 @@ public:
bool outputDimsForwarded() const override final {
std::size_t forwarded = 0;
- for (; forwarded < NUM && (!mInputs[forwarded]->empty()); ++forwarded) {}
- return ((forwarded==NUM) && !(mOutput->empty()));
+ for (; forwarded < mNbInputs && (!mInputs[forwarded]->empty()); ++forwarded) {}
+ return ((forwarded==mNbInputs) && !(mOutput->empty()));
}
// void checkDims() const override final {
@@ -114,13 +112,13 @@ public:
// }
// }
inline Tensor& input(const IOIndex_t inputIdx) const override final {
- assert(static_cast<std::size_t>(inputIdx) < NUM && "wrong inputIdx for Add operator.");
+ assert(static_cast<std::size_t>(inputIdx) < mNbInputs && "wrong inputIdx for Add operator.");
return *(mInputs[inputIdx].get());
}
inline Tensor& output(const IOIndex_t /*outputIdx*/) const override final { return *(mOutput.get()); }
inline std::shared_ptr<Tensor> getInput(const IOIndex_t inputIdx) const override final {
- assert(static_cast<std::size_t>(inputIdx) < NUM && "wrong inputIdx for Add operator.");
+ assert(static_cast<std::size_t>(inputIdx) < mNbInputs && "wrong inputIdx for Add operator.");
return mInputs[inputIdx];
}
inline std::shared_ptr<Tensor> getOutput(const IOIndex_t outputIdx) const override final {
@@ -130,7 +128,7 @@ public:
}
std::shared_ptr<Data> getRawInput(const IOIndex_t inputIdx) const override final {
- assert(static_cast<std::size_t>(inputIdx) < NUM && "wrong inputIdx for Add operator.");
+ assert(static_cast<std::size_t>(inputIdx) < mNbInputs && "wrong inputIdx for Add operator.");
return std::static_pointer_cast<Data>(mInputs[inputIdx]);
}
std::shared_ptr<Data> getRawOutput(const IOIndex_t outputIdx) const override final {
@@ -141,11 +139,11 @@ public:
void setBackend(const std::string& name) override {
- mImpl = Registrar<Add_Op<NUM>>::create(name)(*this);
+ mImpl = Registrar<Add_Op>::create(name)(*this);
mOutput->setBackend(name);
// FIXME: temporary workaround
- for (std::size_t i = 0; i < NUM; ++i) {
+ for (std::size_t i = 0; i < mNbInputs; ++i) {
mInputs[i]->setBackend(name);
}
}
@@ -154,15 +152,16 @@ public:
mOutput->setDatatype(datatype);
// FIXME: temporary workaround
- for (std::size_t i = 0; i < NUM; ++i) {
+ for (std::size_t i = 0; i < mNbInputs; ++i) {
mInputs[i]->setDatatype(datatype);
}
}
- inline IOIndex_t nbInputs() const noexcept override final { return NUM; }
- inline IOIndex_t nbDataInputs() const noexcept override final { return NUM; }
+ inline IOIndex_t nbInputs() const noexcept override final { return mNbInputs; }
+ inline IOIndex_t nbDataInputs() const noexcept override final { return mNbInputs; }
inline IOIndex_t nbOutputs() const noexcept override final { return 1; }
- static const std::vector<std::string> getInputsName(){
+
+ static const std::vector<std::string> getInputsName(){
return {"data_input_0", "data_input_n"};
}
static const std::vector<std::string> getOutputsName(){
@@ -170,9 +169,8 @@ public:
}
};
-template <std::size_t NUM>
-inline std::shared_ptr<Node> Add(const std::string& name = "") {
- return std::make_shared<Node>(std::make_shared<Add_Op<NUM>>(), name);
+inline std::shared_ptr<Node> Add(const IOIndex_t nbIn, const std::string& name = "") {
+ return std::make_shared<Node>(std::make_shared<Add_Op>(nbIn), name);
}
}
diff --git a/python_binding/operator/pybind_Add.cpp b/python_binding/operator/pybind_Add.cpp
index 0b2323c5c..bff795a73 100644
--- a/python_binding/operator/pybind_Add.cpp
+++ b/python_binding/operator/pybind_Add.cpp
@@ -19,15 +19,15 @@
namespace py = pybind11;
namespace Aidge {
-template <std::size_t NUM> void declare_Add(py::module &m) {
- py::class_<Add_Op<NUM>, std::shared_ptr<Add_Op<NUM>>, Operator>(m, "AddOp", py::multiple_inheritance())
- .def("get_inputs_name", &Add_Op<NUM>::getInputsName)
- .def("get_outputs_name", &Add_Op<NUM>::getOutputsName);
+void declare_Add(py::module &m) {
+ py::class_<Add_Op, std::shared_ptr<Add_Op>, Operator>(m, "AddOp", py::multiple_inheritance())
+ .def("get_inputs_name", &Add_Op::getInputsName)
+ .def("get_outputs_name", &Add_Op::getOutputsName);
- m.def("Add", &Add<NUM>, py::arg("name") = "");
+ m.def("Add", &Add, py::arg("nbIn"), py::arg("name") = "");
}
void init_Add(py::module &m) {
- declare_Add<2>(m);
+ declare_Add(m);
}
} // namespace Aidge
diff --git a/unit_tests/recipies/Test_FuseMulAdd.cpp b/unit_tests/recipies/Test_FuseMulAdd.cpp
index 92b2b7c13..7e64f3ff5 100644
--- a/unit_tests/recipies/Test_FuseMulAdd.cpp
+++ b/unit_tests/recipies/Test_FuseMulAdd.cpp
@@ -25,9 +25,9 @@ namespace Aidge {
TEST_CASE("[cpu/recipies] FuseMulAdd", "[FuseMulAdd][recipies]") {
// generate the original GraphView
auto matmul0 = MatMul(5, "matmul0");
- auto add0 = Add<2>("add0");
+ auto add0 = Add(2, "add0");
auto matmul1 = MatMul(5, "matmul1");
- auto add1 = Add<2>("add1");
+ auto add1 = Add(2, "add1");
auto b0 = Producer({5}, "B0");
auto w0 = Producer({5, 5}, "W0");
--
GitLab