feat_operator_convtranspose

Context

issue : aidge#235 (closed)

• ADDED :
  • feat : [ADD] PaddedConvTranspose operator
  • feat : [ADD] convtranspose operator
  • feat : added getDataType() function to OperatorTensor
  • feat : approxEqual, better error checking & error messages - replaced tensor->size by tensor->dims comparison - replaced fmt::print by log::error - replaced assert with AIDGE_ASSERT
• FIXES :
  • fix : Expand shape input was set as Param instead of Data
  • fix : missing header in Types.h
• CHORES :
  • chore : [DOC] added pybind documentation to graphview.compile
• Conv
  • feat : [Conv] added check to ensure dilation & stride attributes values are strictly positive
  • fix : [Conv] added check to ensure that dilation & stride values were all >= 1
    • better warning message for conv operator
  • chore : [Conv] updated the method for attribute accession to ease code reading
  • fix : [Conv] added check to ensure that dilation & stride values were all >= 1 & better warning message for conv operator constructor
  • chore : [Conv] updated the method for attribute accession to ease code reading

include/aidge/utils/TensorUtils.hpp

@@ -11,25 +11,23 @@
 
 #ifndef AIDGE_CORE_UTILS_TENSOR_UTILS_H_
 #define AIDGE_CORE_UTILS_TENSOR_UTILS_H_
-
-#include <cmath>  // std::abs
-
-#include "aidge/data/DataType.hpp"
 #include "aidge/data/Tensor.hpp"
-#include "aidge/utils/ErrorHandling.hpp"
-#include "aidge/utils/Log.hpp"
+#include <cmath> // std::abs
+#include <fmt/base.h>
 
 namespace Aidge {
 
 /**
- * @brief Compare two Aidge::Tensor value wise. The comparison function is:
+ * @brief Compare two :cpp:class:`Aidge::Tensor` value wise. The comparison
+ * function is:
  *
  * |t1-t2| <= absolute + relative * |t2|
  *
  * If a tensor value is different from the other tensor return False
  * If the tensor does not have the same size, return False
  * If the datatype is not the same between each tensor return False
- * If the templated type does not correspond to the datatype of each tensor, raise an assertion error
+ * If the templated type does not correspond to the datatype of each tensor,
+ * raise an assertion error
  *
  * @tparam T1 should correspond to the type of the first tensor, defines part of the type for absolute and relative error
  * @tparam T2 should correspond to the type of the second tensor, defaults to T1
@@ -43,33 +41,42 @@ template <typename T1, typename T2 = T1>
 bool approxEq(const Tensor& t1, const Tensor& t2, float relative = 1e-5f, float absolute = 1e-8f) {
     // Check template type matches tensor datatype
     if (t1.dataType() != NativeType_v<T1>) {
-        Log::error("First tensor datatype ({}) does not match template type", t1.dataType());
+        Log::error("approxEq : First tensor datatype ({}) does not match"
+                   "template type (NativeType_v<T1> = {}) .",
+                    t1.dataType(),
+                    NativeType_v<T1>);
         return false;
     }
 
     if (t2.dataType() != NativeType_v<T2>) {
-        Log::error("Second tensor datatype ({}) does not match template type", t2.dataType());
+        Log::error("approxEq : Second tensor datatype ({}) does not match"
+                   "template type (NativeType_v<T1> = {}) .",
+                    t2.dataType(),
+                    NativeType_v<T1>);
         return false;
     }
 
     // Validate parameters
     if (relative < 0.0f) {
-        Log::error("Relative error must be non-negative (got {})", relative);
+        Log::error("approxEq : Relative error must be non-negative (got  : {}).",
+                    relative);
         return false;
     }
 
     if (absolute < 0.0f || absolute > 1.0f) {
-        Log::error("Absolute error must be between 0 and 1 (got {})", absolute);
+        Log::error("approxEq : Absolute error must be between 0 and 1 (got  : {}).",
+                    absolute);
         return false;
     }
 
-    // Check tensor sizes match
-    if (t1.size() != t2.size()) {
-        Log::error("Tensor sizes do not match: {} vs {}", t1.size(), t2.size());
+    if (t1.dims() != t2.dims()) {
+        Log::error("approxEq: Dimension mismatch.\nt1 :\t{}\nt2 :\t{}",
+                   t1.dims(),
+                   t2.dims());
         return false;
     }
 
-    // Compare values/
+    // Compare values
     for (size_t i = 0; i < t1.size(); ++i) {
         const auto val1 = t1.get<T1>(i);
         const auto val2 = t2.get<T2>(i);
@@ -77,9 +84,15 @@ bool approxEq(const Tensor& t1, const Tensor& t2, float relative = 1e-5f, float
         const float threshold = absolute + (relative * static_cast<float>(std::abs(val2)));
 
         if (diff > threshold) {
-            Log::notice("Tensor values differ at index {}: {} vs {} (diff: {}, threshold: {})\n"
-                "Tensor 1:\n{}\nTensor 2:\n{}",
-                i, val1, val2, diff, threshold, t1, t2);
+            Log::error("approxEq : value mismatch at index {} : {} != "
+                       "{} (diff: {}, threshold: {}) \nt1:\n{}\nt2:\n{}\n",
+                       i,
+                       t1.get<T1>(i),
+                       t2.get<T1>(i),
+                       diff,
+                       threshold,
+                       t1,
+                       t2);
             return false;
         }
     }