[Eager] Support transformer tests in eager mode (#41347)

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

@@ -21,7 +21,7 @@ from paddle.fluid import Embedding, LayerNorm, Linear, Layer
 from paddle.fluid.dygraph import to_variable, guard
 from paddle.fluid.dygraph import TracedLayer
 from test_imperative_base import new_program_scope
-from paddle.fluid.framework import _test_eager_guard
+from paddle.fluid.framework import _test_eager_guard, in_dygraph_mode, _in_legacy_dygraph
 from paddle.fluid import core
 import numpy as np
 import six
@@ -1041,6 +1041,7 @@ class TestDygraphTransformerSortGradient(unittest.TestCase):
 
         with guard():
             fluid.set_flags({'FLAGS_sort_sum_gradient': True})
+            if _in_legacy_dygraph():
                 dy_avg_cost_value, dy_sum_cost_value, dy_predict_value, dy_token_num_value, \
                     dy_param_init, dy_param_updated = run_dygraph()
 
@@ -1116,12 +1117,13 @@ class TestDygraphTransformerSortGradient(unittest.TestCase):
                     for k in range(4, len(out)):
                         static_param_updated[static_param_name_list[k -
                                                                     4]] = out[k]
-
+        if _in_legacy_dygraph():
             self.assertTrue(
                 np.array_equal(static_avg_cost_value, dy_avg_cost_value))
             self.assertTrue(
                 np.array_equal(static_sum_cost_value, dy_sum_cost_value))
-        self.assertTrue(np.array_equal(static_predict_value, dy_predict_value))
+            self.assertTrue(
+                np.array_equal(static_predict_value, dy_predict_value))
             self.assertTrue(
                 np.array_equal(static_token_num_value, dy_token_num_value))
 
@@ -1130,7 +1132,7 @@ class TestDygraphTransformerSortGradient(unittest.TestCase):
             for key, value in six.iteritems(static_param_updated):
                 self.assertTrue(np.array_equal(value, dy_param_updated[key]))
 
-        # check eager result
+        # compare eager result with imperative result
         with guard():
             fluid.set_flags({'FLAGS_sort_sum_gradient': False})
             dy_avg_cost_value, dy_sum_cost_value, dy_predict_value, dy_token_num_value, \