Created by: Banmahhhh

• 版本、环境信息：    1）PaddlePaddle版本：paddlepaddle1.5.1    2）CPU：2.7 GHz Intel Core i5    3）系统环境：Mac OS 10.14，Python3.6.9
• 代码：

        act_prob = self.model(obs)
        action = fluid.dygraph.to_variable(action)
        reward = fluid.dygraph.to_variable(reward)

        log_prob = layers.cross_entropy(act_prob, action)
        cost = log_prob * reward
        cost = layers.cast(cost, dtype='float32')
        cost = layers.reduce_mean(cost)
        cost.backward()
        self.optimizer.minimize(cost)
        self.model.clear_gradients()

• error message：

  File "policy_gradient.py", line 28, in learn
    self.optimizer.minimize(cost)
  File "</anaconda3/envs/*****/lib/python3.6/site-packages/decorator.py:decorator-gen-20>", line 2, in minimize
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in __impl__
    return wrapped_func(*args, **kwargs)
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/dygraph/base.py", line 87, in __impl__
    return func(*args, **kwargs)
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/optimizer.py", line 601, in minimize
    loss, startup_program=startup_program, params_grads=params_grads)
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/optimizer.py", line 558, in apply_optimize
    optimize_ops = self._create_optimization_pass(params_grads)
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/optimizer.py", line 376, in _create_optimization_pass
    param_and_grad)
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/optimizer.py", line 1383, in _append_optimize_op
    stop_gradient=True)
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/framework.py", line 1762, in append_op
    kwargs.get("stop_gradient", False))
  File "/anaconda3/envs/*****/lib/python3.6/site-packages/paddle/fluid/dygraph/tracer.py", line 59, in trace_op
    framework._current_expected_place(), stop_gradient)
paddle.fluid.core_avx.EnforceNotMet: holder_ should not be null