Python unitest "test_network_with_dtype" keeps failing.

[wanglei828]

When running "test_all_ops", the "test_network_with_dtype" keeps failing. The error message as follows:

ERROR: test_cpu (test_network_with_dtype.TestNetWithDtype)

Traceback (most recent call last):
File "test_network_with_dtype.py", line 58, in test_cpu
self.set_network()
File "test_network_with_dtype.py", line 30, in set_network
self.x = fluid.layers.data(name='x', shape=[13], dtype=self.dtype)
File "/paddle/build/python/paddle/fluid/layers/io.py", line 82, in data
is_data=True)
File "/paddle/build/python/paddle/fluid/layer_helper.py", line 344, in create_global_variable
*args, persistable=persistable, **kwargs)
File "/paddle/build/python/paddle/fluid/framework.py", line 763, in create_var
var = Variable(block=self, *args, **kwargs)
File "/paddle/build/python/paddle/fluid/framework.py", line 195, in init
"matched.".format(self.name, old_shape, shape))
ValueError: Variable x has been created before. the previous shape is (-1L, 3L); the new shape is (-1, 13). They are not matched.

======================================================================
ERROR: test_gpu (test_network_with_dtype.TestNetWithDtype)

Traceback (most recent call last):
File "test_network_with_dtype.py", line 65, in test_gpu
self.set_network()
File "test_network_with_dtype.py", line 30, in set_network
self.x = fluid.layers.data(name='x', shape=[13], dtype=self.dtype)
File "/paddle/build/python/paddle/fluid/layers/io.py", line 82, in data
is_data=True)
File "/paddle/build/python/paddle/fluid/layer_helper.py", line 344, in create_global_variable
*args, persistable=persistable, **kwargs)
File "/paddle/build/python/paddle/fluid/framework.py", line 763, in create_var
var = Variable(block=self, *args, **kwargs)
File "/paddle/build/python/paddle/fluid/framework.py", line 195, in init
"matched.".format(self.name, old_shape, shape))
ValueError: Variable x has been created before. the previous shape is (-1L, 3L); the new shape is (-1, 13). They are not matched.

[wanglei828]

Hi, I don't think this has been fixed. In CI and local dev, we can still see this test failure.
The error message as follows:

ERROR: test_cpu (test_network_with_dtype.TestNetWithDtype)

Traceback (most recent call last):
File "test_network_with_dtype.py", line 63, in test_cpu
self.run_net_on_place(place)
File "test_network_with_dtype.py", line 49, in run_net_on_place
exe.run(fluid.default_startup_program())
File "/paddle/build/python/paddle/fluid/executor.py", line 337, in run
fetch_var_name=fetch_var_name)
File "/paddle/build/python/paddle/fluid/executor.py", line 209, in _add_feed_fetch_ops
tmp_program = program.clone()
File "/paddle/build/python/paddle/fluid/framework.py", line 1055, in clone
p.sync_with_cpp()
File "/paddle/build/python/paddle/fluid/framework.py", line 1160, in sync_with_cpp
block.sync_with_cpp()
File "/paddle/build/python/paddle/fluid/framework.py", line 909, in sync_with_cpp
op = Operator(self, op_desc)
File "/paddle/build/python/paddle/fluid/framework.py", line 409, in init
"type to initilized an Operator can not be None.")
ValueError: type to initilized an Operator can not be None.

======================================================================
ERROR: test_gpu (test_network_with_dtype.TestNetWithDtype)

Traceback (most recent call last):
File "test_network_with_dtype.py", line 70, in test_gpu
self.run_net_on_place(place)
File "test_network_with_dtype.py", line 49, in run_net_on_place
exe.run(fluid.default_startup_program())
File "/paddle/build/python/paddle/fluid/executor.py", line 337, in run
fetch_var_name=fetch_var_name)
File "/paddle/build/python/paddle/fluid/executor.py", line 209, in _add_feed_fetch_ops
tmp_program = program.clone()
File "/paddle/build/python/paddle/fluid/framework.py", line 1055, in clone
p.sync_with_cpp()
File "/paddle/build/python/paddle/fluid/framework.py", line 1160, in sync_with_cpp
block.sync_with_cpp()
File "/paddle/build/python/paddle/fluid/framework.py", line 909, in sync_with_cpp
op = Operator(self, op_desc)
File "/paddle/build/python/paddle/fluid/framework.py", line 409, in init
"type to initilized an Operator can not be None.")
ValueError: type to initilized an Operator can not be None.