【Eager】fix multiply double grad error (#52870)

* fix multiply double grad error

* fix multiply dy only kenrel

paddle/phi/kernels/impl/elementwise_grad_kernel_impl.h

@@ -529,6 +529,44 @@ void MultiplyDoubleGradKernel(const Context& dev_ctx,
                                           funcs::InverseMultiplyFunctor<T>>(
             dev_ctx, dout, ddy_safe, dx, axis);
 
+      } else if ((!dx) && dy) {
+        DenseTensor tmp_a(ddout->dtype());
+        tmp_a.Resize(ddout->dims());
+
+        dev_ctx.template Alloc<T>(&tmp_a);
+        funcs::DefaultElementwiseOperator<Context,
+                                          T,
+                                          funcs::MultiplyFunctor<T>,
+                                          funcs::InverseMultiplyFunctor<T>>(
+            dev_ctx, x, ddy_safe, &tmp_a, axis);
+
+        auto ddout_t1 = phi::EigenVector<T>::Flatten(tmp_a);
+
+        funcs::DefaultElementwiseOperator<Context,
+                                          T,
+                                          funcs::MultiplyFunctor<T>,
+                                          funcs::InverseMultiplyFunctor<T>>(
+            dev_ctx, ddx_safe, y, ddout, axis);
+
+        auto ddout_t2 = phi::EigenVector<T>::Flatten(*ddout);
+        ddout_t2.device(place) = ddout_t2 + ddout_t1;
+
+        // NOTE: in the following ElemwiseGradCompute, for the
+        // first output tensor is nullptr, the branch to calculate first
+        // output tensor will not be activated, DivGradDx function will not
+        // be called and can be ignored, the first branch has little effect
+        // on running speed.
+        phi::funcs::ElemwiseGradCompute<Context, T, MulGradDX<T>, MulGradDY<T>>(
+            dev_ctx,
+            ddx_safe,
+            ddy_safe,
+            dout,
+            dout,
+            axis,
+            nullptr,
+            dy,
+            MulGradDX<T>(),
+            MulGradDY<T>());
       } else {
         DenseTensor tmp_a(ddout->dtype());
         tmp_a.Resize(ddout->dims());
@@ -554,7 +592,7 @@ void MultiplyDoubleGradKernel(const Context& dev_ctx,
       }
     }
   } else {
-    if (dx && dy) {
+    VLOG(3) << "Calculating here with dx: " << dx << ", dy: " << dy;
     phi::funcs::ElemwiseGradCompute<Context, T, MulGradDX<T>, MulGradDY<T>>(
         dev_ctx,
         ddx_safe,
@@ -567,7 +605,6 @@ void MultiplyDoubleGradKernel(const Context& dev_ctx,
         MulGradDX<T>(),
         MulGradDY<T>());
   }
-  }
 }
 
 template <typename T, typename Context>