[Eager] Support gnn ptb_rnn in eager mode (#39993)

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

@@ -23,6 +23,7 @@ import paddle.fluid.core as core
 from paddle.fluid.optimizer import AdamOptimizer
 from test_imperative_base import new_program_scope
 from paddle.fluid.dygraph.base import to_variable
+from paddle.fluid.framework import _test_eager_guard
 
 
 def gen_data():
@@ -60,7 +61,7 @@ class GCN(fluid.Layer):
 
 
 class TestDygraphGNN(unittest.TestCase):
-    def test_gnn_float32(self):
+    def func_gnn_float32(self):
         paddle.seed(90)
         paddle.framework.random._manual_program_seed(90)
         startup = fluid.Program()
@@ -168,6 +169,11 @@ class TestDygraphGNN(unittest.TestCase):
         self.assertTrue(np.allclose(static_weight, model2_gc_weight_value))
         sys.stderr.write('%s %s\n' % (static_loss, loss_value))
 
+    def test_gnn_float32(self):
+        with _test_eager_guard():
+            self.func_gnn_float32()
+        self.func_gnn_float32()
+
 
 if __name__ == '__main__':
     unittest.main()

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

@@ -26,10 +26,11 @@ from test_imperative_base import new_program_scope
 from test_imperative_ptb_rnn import PtbModel
 import numpy as np
 import six
+from paddle.fluid.framework import _test_eager_guard
 
 
 class TestDygraphPtbRnnSortGradient(unittest.TestCase):
-    def test_ptb_rnn_sort_gradient(self):
+    def func_ptb_rnn_sort_gradient(self):
         for is_sparse in [True, False]:
             self.ptb_rnn_sort_gradient_cpu_float32(is_sparse)
 
@@ -171,6 +172,11 @@ class TestDygraphPtbRnnSortGradient(unittest.TestCase):
         for key, value in six.iteritems(static_param_updated):
             self.assertTrue(np.array_equal(value, dy_param_updated[key]))
 
+    def test_ptb_rnn_sort_gradient(self):
+        with _test_eager_guard():
+            self.func_ptb_rnn_sort_gradient()
+        self.func_ptb_rnn_sort_gradient()
+
 
 if __name__ == '__main__':
     unittest.main()