diff --git a/include/aidge/backend/cpu/data/TensorImpl.hpp b/include/aidge/backend/cpu/data/TensorImpl.hpp
index 9e1f0335edd47ae1405fc9f9949340c13a67221f..e9ad19291cc3907b3ef4a92370a3bafcabf37545 100644
--- a/include/aidge/backend/cpu/data/TensorImpl.hpp
+++ b/include/aidge/backend/cpu/data/TensorImpl.hpp
@@ -51,16 +51,16 @@ public:
         const T* srcT = static_cast<const T *>(src);
         T* dstT = static_cast<T *>(rawPtr(offset));
 
-        AIDGE_ASSERT(length <= mData.size() || length <= mNbElts, "copy length is above capacity");
-        AIDGE_ASSERT(dstT < srcT || dstT >= srcT + length, "overlapping copy is not supported");
+        AIDGE_ASSERT(length <= mData.size() || length <= mNbElts, "TensorImpl_cpu<{}>::copy(): copy length ({}) is above capacity ({})", typeid(T).name(), length, mNbElts);
+        AIDGE_ASSERT(dstT < srcT || dstT >= srcT + length, "TensorImpl_cpu<{}>::copy(): overlapping copy is not supported", typeid(T).name());
         std::copy(srcT, srcT + length, dstT);
     }
 
     void copyCast(const void *src, const DataType srcDt, NbElts_t length, NbElts_t offset = 0) override final;
 
     void copyFromDevice(const void *src, const std::pair<std::string, DeviceIdx_t>& device, NbElts_t length, NbElts_t offset = 0) override final {
-        AIDGE_ASSERT(device.first == Backend, "backend must match");
-        AIDGE_ASSERT(device.second == 0, "device cannot be != 0 for CPU backend");
+        AIDGE_ASSERT(device.first == Backend, "TensorImpl_cpu<{}>::copyFromDevice(): backend must match", typeid(T).name());
+        AIDGE_ASSERT(device.second == 0, "TensorImpl_cpu<{}>::copyFromDevice(): device ({}) cannot be != 0 for CPU backend", typeid(T).name(), device.second);
         copy(src, length, offset);
     }
 
@@ -70,7 +70,7 @@ public:
 
     void copyToHost(void *dst, NbElts_t length, NbElts_t offset = 0) const override final {
         const T* src = static_cast<const T*>(rawPtr(offset));
-        AIDGE_ASSERT(length <= mData.size() || length <= mNbElts, "copy length is above capacity");
+        AIDGE_ASSERT(length <= mData.size() || length <= mNbElts, "TensorImpl_cpu<{}>::copyToHost(): copy length ({}) is above capacity ({})", typeid(T).name(), length, mNbElts);
         std::copy(src, src + length, static_cast<T *>(dst));
     }
 
@@ -80,7 +80,7 @@ public:
     };
 
     const void *rawPtr(NbElts_t offset = 0) const override final {
-        AIDGE_ASSERT(mData.size() >= mNbElts, "accessing uninitialized const rawPtr");
+        AIDGE_ASSERT(mData.size() >= mNbElts, "TensorImpl_cpu<{}>::rawPtr(): accessing uninitialized const rawPtr", typeid(T).name());
         return (mData.data() + offset);
     };
 
@@ -90,12 +90,12 @@ public:
     };
 
     const void *hostPtr(NbElts_t offset = 0) const override final {
-        AIDGE_ASSERT(mData.size() >= mNbElts, "accessing uninitialized const hostPtr");
+        AIDGE_ASSERT(mData.size() >= mNbElts, "TensorImpl_cpu<{}>::hostPtr(): accessing uninitialized const hostPtr", typeid(T).name());
         return (mData.data() + offset);
     };
 
     void setRawPtr(void *ptr, NbElts_t length) override final {
-        AIDGE_ASSERT(length >= mNbElts, "trying to set raw pointer of insufficient capacity");
+        AIDGE_ASSERT(length >= mNbElts, "TensorImpl_cpu<{}>::setRawPtr(): trying to set raw pointer (length: {}) of insufficient capacity (required: {})", typeid(T).name(), length, mNbElts);
         mData = future_std::span<T>(static_cast<T *>(ptr), length);
         mDataOwner.reset();
     };
@@ -106,7 +106,7 @@ private:
     void lazyInit() {
         if (mData.size() < mNbElts) {
             // Need more data, a re-allocation will occur
-            AIDGE_ASSERT(mData.empty() || mDataOwner != nullptr, "trying to enlarge non-owned data");
+            AIDGE_ASSERT(mData.empty() || mDataOwner != nullptr, "TensorImpl_cpu<{}>: trying to enlarge non-owned data", typeid(T).name());
             mDataOwner.reset(new T[mNbElts]);
             mData = future_std::span<T>(mDataOwner.get(), mNbElts);
         }
diff --git a/src/backend/cpu/data/TensorImpl.cpp b/src/backend/cpu/data/TensorImpl.cpp
index da90197e912fbeabc2f28bd3bedd91cc6f29e466..39ca71de9ea1bdb89f8138c3e38a016b516686dd 100644
--- a/src/backend/cpu/data/TensorImpl.cpp
+++ b/src/backend/cpu/data/TensorImpl.cpp
@@ -53,7 +53,7 @@ void Aidge::TensorImpl_cpu<T>::copyCast(const void *src, const Aidge::DataType s
     }
 
     T* dstT = static_cast<T *>(rawPtr(offset));
-    AIDGE_ASSERT(length <= mData.size() || length <= mNbElts, "copy length is above capacity");
+    AIDGE_ASSERT(length <= mData.size() || length <= mNbElts, "TensorImpl_cpu<{}>::copyCast(): copy length ({}) is above capacity ({})", typeid(T).name(), length, mNbElts);
     switch (srcDt)
     {
         case DataType::Float64: