This unit test error on CI has encountered many times. [test_network_with_dtype]

[chengduoZH]

[13:45:25] :	 [Step 1/1] test_cpu (test_network_with_dtype.TestNetWithDtype) ... ERROR
[13:45:25] :	 [Step 1/1] test_gpu (test_network_with_dtype.TestNetWithDtype) ... ERROR
[13:45:25] :	 [Step 1/1] 
[13:45:25] :	 [Step 1/1] ======================================================================
[13:45:25] :	 [Step 1/1] ERROR: test_cpu (test_network_with_dtype.TestNetWithDtype)
[13:45:25] :	 [Step 1/1] ----------------------------------------------------------------------
[13:45:25] :	 [Step 1/1] Traceback (most recent call last):
[13:45:25] :	 [Step 1/1]   File "test_network_with_dtype.py", line 63, in test_cpu
[13:45:25] :	 [Step 1/1]     self.run_net_on_place(place)
[13:45:25] :	 [Step 1/1]   File "test_network_with_dtype.py", line 49, in run_net_on_place
[13:45:25] :	 [Step 1/1]     exe.run(fluid.default_startup_program())
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/executor.py", line 337, in run
[13:45:25] :	 [Step 1/1]     fetch_var_name=fetch_var_name)
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/executor.py", line 209, in _add_feed_fetch_ops
[13:45:25] :	 [Step 1/1]     tmp_program = program.clone()
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 1055, in clone
[13:45:25] :	 [Step 1/1]     p.sync_with_cpp()
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 1160, in sync_with_cpp
[13:45:25] :	 [Step 1/1]     block.sync_with_cpp()
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 909, in sync_with_cpp
[13:45:25] :	 [Step 1/1]     op = Operator(self, op_desc)
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 409, in __init__
[13:45:25] :	 [Step 1/1]     "`type` to initilized an Operator can not be None.")
[13:45:25] :	 [Step 1/1] ValueError: `type` to initilized an Operator can not be None.
[13:45:25] :	 [Step 1/1] 
[13:45:25] :	 [Step 1/1] ======================================================================
[13:45:25] :	 [Step 1/1] ERROR: test_gpu (test_network_with_dtype.TestNetWithDtype)
[13:45:25] :	 [Step 1/1] ----------------------------------------------------------------------
[13:45:25] :	 [Step 1/1] Traceback (most recent call last):
[13:45:25] :	 [Step 1/1]   File "test_network_with_dtype.py", line 70, in test_gpu
[13:45:25] :	 [Step 1/1]     self.run_net_on_place(place)
[13:45:25] :	 [Step 1/1]   File "test_network_with_dtype.py", line 49, in run_net_on_place
[13:45:25] :	 [Step 1/1]     exe.run(fluid.default_startup_program())
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/executor.py", line 337, in run
[13:45:25] :	 [Step 1/1]     fetch_var_name=fetch_var_name)
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/executor.py", line 209, in _add_feed_fetch_ops
[13:45:25] :	 [Step 1/1]     tmp_program = program.clone()
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 1055, in clone
[13:45:25] :	 [Step 1/1]     p.sync_with_cpp()
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 1160, in sync_with_cpp
[13:45:25] :	 [Step 1/1]     block.sync_with_cpp()
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 909, in sync_with_cpp
[13:45:25] :	 [Step 1/1]     op = Operator(self, op_desc)
[13:45:25] :	 [Step 1/1]   File "/paddle/build/python/paddle/fluid/framework.py", line 409, in __init__
[13:45:25] :	 [Step 1/1]     "`type` to initilized an Operator can not be None.")
[13:45:25] :	 [Step 1/1] ValueError: `type` to initilized an Operator can not be None.
[13:45:25] :	 [Step 1/1] 
[13:45:25] :	 [Step 1/1] ----------------------------------------------------------------------
[13:45:25] :	 [Step 1/1] Ran 1677 tests in 189.090s
[13:45:25] :	 [Step 1/1] 
[13:45:25] :	 [Step 1/1] FAILED (errors=2)
[13:45:25] :	 [Step 1/1]