!15209 fix l2normalize ops parallel problem

From: @yuanwei66
Reviewed-by: @wuxuejian,@c_34
Signed-off-by: @wuxuejian,@c_34

mindspore/ccsrc/backend/kernel_compiler/cpu/l2_normalize_cpu_kernel.cc

@@ -22,7 +22,7 @@ namespace kernel {
 template <typename T>
 void L2NormalizeCPUKernel<T>::InitKernel(const CNodePtr &kernel_node) {
   MS_EXCEPTION_IF_NULL(kernel_node);
-  epsilon_ = AnfAlgo::GetNodeAttr<float>(kernel_node, "epsilon");
+  epsilon_ = static_cast<T>(AnfAlgo::GetNodeAttr<float>(kernel_node, "epsilon"));
   axis_ = LongToInt(AnfAlgo::GetNodeAttr<int64_t>(kernel_node, "axis"));
   input_shape_ = AnfAlgo::GetInputDeviceShape(kernel_node, 0);
   output_shape_ = AnfAlgo::GetOutputInferShape(kernel_node, 0);
@@ -35,9 +35,6 @@ void L2NormalizeCPUKernel<T>::InitKernel(const CNodePtr &kernel_node) {
 template <typename T>
 void L2NormalizeCPUKernel<T>::CalcDenominator(const T *input_addr, const size_t reduce_size, const int dims,
                                               std::unique_ptr<T[]> *denominator_addr) {
-  T temp = (T)0.0;
-  T epsilon = (T)epsilon_;
-  T denominator = (T)0.0;
   // Calculate transpose axes and stride
   size_t stride = 1;
   std::vector<size_t> axes(input_shape_.size());
@@ -60,6 +57,8 @@ void L2NormalizeCPUKernel<T>::CalcDenominator(const T *input_addr, const size_t
   TransposeIterator tran_base_iter(std::move(transpose_shape), std::move(axes), input_shape_);
 
   auto task = [&](size_t start, size_t end) {
+    T temp = (T)0.0;
+    T denominator = (T)0.0;
     auto iter = tran_base_iter;
     iter.SetPos(start * stride);
     for (size_t i = start; i < end; ++i) {
@@ -71,7 +70,7 @@ void L2NormalizeCPUKernel<T>::CalcDenominator(const T *input_addr, const size_t
         denominator += temp * temp;
         iter.GenNextPos();
       }
-      denominator = (denominator > epsilon) ? denominator : epsilon;
+      denominator = (denominator > epsilon_) ? denominator : epsilon_;
       (*denominator_addr)[i] = sqrt(denominator);
     }
   };
@@ -146,7 +145,7 @@ void L2NormalizeCPUKernel<T>::CheckParam(const CNodePtr &kernel_node) {
   if (axis_ < -dims || axis_ >= dims) {
     MS_LOG(EXCEPTION) << "Attr axis_ " << axis_ << " must be in " << -dims << "~" << dims;
   }
-  if (epsilon_ == 0.0) {
+  if (epsilon_ == (T)0.0) {
     MS_LOG(EXCEPTION) << "Attr epsilon can not be zero.";
   }
 }

mindspore/ccsrc/backend/kernel_compiler/cpu/l2_normalize_cpu_kernel.h

@@ -45,7 +45,7 @@ class L2NormalizeCPUKernel : public CPUKernel {
  private:
   std::vector<size_t> input_shape_;
   std::vector<size_t> output_shape_;
-  float epsilon_;
+  T epsilon_;
   int axis_;
   void CheckParam(const CNodePtr &kernel_node);
 };

tests/st/ops/cpu/test_l2normalize_op.py

@@ -37,10 +37,10 @@ class Net(Cell):
 @pytest.mark.platform_x86_cpu_training
 @pytest.mark.env_onecard
 def test_l2normalize_float32():
-    x = np.arange(96).astype(np.float32).reshape(2, 3, 4, 4)
+    x = np.arange(20*20*20*20).astype(np.float32).reshape(20, 20, 20, 20)
     expect = x / np.sqrt(np.sum(x**2, axis=0, keepdims=True))
     x = Tensor(x)
-    error = np.ones(shape=[2, 3, 4, 4]) * 1.0e-5
+    error = np.ones(shape=[20, 20, 20, 20]) * 1.0e-5
 
     norm_op = Net(axis=0)
     output = norm_op(x)