Fixed CuDNN descriptor lazy initialization

include/aidge/backend/cuda/operator/ConvImpl.hpp

@@ -44,10 +44,7 @@ private:
     void* mWorkspace = nullptr;
 
 public:
-    ConvImpl_cuda(const Conv_Op<DIM> &op) : mOp(op) {
-        CHECK_CUDNN_STATUS(cudnnCreateConvolutionDescriptor(&mConvDesc));
-        CHECK_CUDNN_STATUS(cudnnCreateFilterDescriptor(&mFilterDesc));
-    }
+    ConvImpl_cuda(const Conv_Op<DIM> &op) : mOp(op) {}
 
     static std::unique_ptr<ConvImpl_cuda> create(const Conv_Op<2> &op) {
         return std::make_unique<ConvImpl_cuda>(op);

src/operator/ConvImpl.cpp

@@ -78,12 +78,13 @@ void Aidge::ConvImpl_cuda<DIM>::forward() {
     assert(mOp.getInput(0) && "missing input #0");
     assert(mOp.getInput(1) && "missing input #1");
 
-    // Initialize CuDNN convolution descriptor
+    // Lazy-initialize CuDNN convolution descriptor
     if (mConvDesc == nullptr) {
         const std::vector<int> strides(mOp.template get<ConvParam::StrideDims>().begin(), mOp.template get<ConvParam::StrideDims>().end());
         const std::vector<int> paddings(DIM, 0);
         const std::vector<int> upscales(mOp.template get<ConvParam::DilationDims>().begin(), mOp.template get<ConvParam::DilationDims>().end());
 
+        CHECK_CUDNN_STATUS(cudnnCreateConvolutionDescriptor(&mConvDesc));
         CHECK_CUDNN_STATUS(
             cudnnSetConvolutionNdDescriptor(mConvDesc,
                                             DIM,
@@ -94,10 +95,11 @@ void Aidge::ConvImpl_cuda<DIM>::forward() {
                                             DataTypeToCudnn(mOp.getOutput(0)->dataType())));
     }
 
-    // Initialize CuDNN filter descriptor
+    // Lazy-initialize CuDNN filter descriptor
     if (mFilterDesc == nullptr) {
         const std::vector<int> kernels(mOp.getInput(1)->dims().begin(), mOp.getInput(1)->dims().end());
 
+        CHECK_CUDNN_STATUS(cudnnCreateFilterDescriptor(&mFilterDesc));
         CHECK_CUDNN_STATUS(cudnnSetFilterNdDescriptor(mFilterDesc,
                                                     DataTypeToCudnn(mOp.getInput(1)->dataType()),
                                                     CUDNN_TENSOR_NCHW,