Support backward during backward.

python/paddle/autograd/py_layer.py

@@ -176,8 +176,9 @@ def backward(ctx, dy):
 
 class PyLayerBackward(PyLayerContext):
     def backward(self, *args, **kwargs):
-        with paddle.fluid.dygraph.no_grad():
-            return self._forward_cls.backward(*args, **kwargs)
+        with paddle.fluid.dygraph.guard():
+            with paddle.fluid.dygraph.no_grad():
+                return self._forward_cls.backward(*args, **kwargs)
 
 
 class LayerMeta(type):

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

@@ -283,20 +283,54 @@ def test_pylayer_inplace(self):
         class cus_tanh(PyLayer):
             @staticmethod
             def forward(ctx, x):
-                return x.mean()
+                return x
 
             @staticmethod
             def backward(ctx, dy):
                 return dy
 
+        class Layer(paddle.nn.Layer):
+            def __init__(self):
+                super(Layer, self).__init__()
+
+            def forward(self, data):
+                data = paddle.nn.functional.relu(data)
+                z = paddle.tanh(data)
+                z = cus_tanh.apply(data)
+                return z.mean()
+
         for i in range(2):
             data = paddle.ones([2, 3], dtype="float64") / (i + 1)
             data.stop_gradient = False
+            layer = Layer()
+            z = layer(data)
+            z.backward()
+            self.assertTrue(data.grad is not None)
+
+    def test_backward_in_backward(self):
+        class cus_tanh(PyLayer):
+            @staticmethod
+            def forward(ctx, x):
+                temp = x.detach()
+                ctx.inputs = temp
+                return x.mean()
+
+            @staticmethod
+            def backward(ctx, dy):
+                with paddle.set_grad_enabled(True):
+                    temp = ctx.inputs
+                    temp.stop_gradient = False
+                    z = paddle.tanh(temp)
+                    z.backward()
+                    self.assertTrue(temp.grad is not None)
+                    return paddle.to_tensor(temp.grad)
+
+        for i in range(2):
+            data = paddle.ones([2, 3], dtype="float32") / (i + 1)
+            data.stop_gradient = False
             data = paddle.nn.functional.relu(data)
             z = paddle.tanh(data)
             z = cus_tanh.apply(data)
-            z.backward()
-            self.assertTrue(data.grad is not None)
 
 
 if __name__ == '__main__':