fix reduce_max/reduce_min bug (#38478)

paddle/fluid/operators/reduce_ops/reduce_op.h

@@ -241,11 +241,12 @@ class ReduceKernel : public framework::OpKernel<T> {
     framework::proto::VarType::Type cast_out_dtype;
 
     // The dims has full dim, set the reduce_all is True
-    const auto& input_dim_size = context.Input<Tensor>("X")->dims().size();
+    const int& input_dim_size = context.Input<Tensor>("X")->dims().size();
     std::set<int> dims_set(dims.begin(), dims.end());
     bool full_dim = true;
-    for (auto i = 0; i < input_dim_size; i++) {
-      if (dims_set.find(i) == dims_set.end()) {
+    for (int i = 0; i < input_dim_size; i++) {
+      if (dims_set.find(i) == dims_set.end() &&
+          dims_set.find(i - input_dim_size) == dims_set.end()) {
         full_dim = false;
         break;
       }

python/paddle/fluid/tests/unittests/test_max_op.py

@@ -98,6 +98,15 @@ class ApiMaxTest(unittest.TestCase):
         self.assertEqual((np_z1 == z_expected).all(), True)
         self.assertEqual((np_z2 == z_expected).all(), True)
 
+    def test_all_negative_axis(self):
+        paddle.disable_static()
+        x = paddle.rand(shape=[2, 2])
+        np_x = x.numpy()
+        z1 = paddle.max(x, axis=(-2, -1))
+        np_z1 = z1.numpy()
+        z_expected = np.array(np.max(np_x, axis=(0, 1)))
+        self.assertEqual((np_z1 == z_expected).all(), True)
+
 
 class TestOutDtype(unittest.TestCase):
     def test_max(self):