提交 de469d58 编写于 作者: Q qiaolongfei

optimize test

上级 d89e1449
...@@ -22,44 +22,55 @@ import paddle.v2.fluid as fluid ...@@ -22,44 +22,55 @@ import paddle.v2.fluid as fluid
class TestPythonOperatorOverride(unittest.TestCase): class TestPythonOperatorOverride(unittest.TestCase):
def check_result(self, fn, x_val, y_val, place, dtype): def check_result(self, fn, place, dtype):
shape = [9, 10] shape = [9, 10]
x_data = np.full(shape, x_val).astype(dtype) x_data = np.random.random(size=shape).astype(dtype)
y_data = np.full(shape, y_val).astype(dtype) y_data = np.random.random(size=shape).astype(dtype)
python_out = fn(x_data, y_data) python_out = fn(x_data, y_data)
x_var = layers.create_global_var( x_var = layers.create_global_var(
shape=shape, value=x_val, dtype=dtype, persistable=True) name='x', shape=shape, value=0.0, dtype=dtype, persistable=True)
y_var = layers.create_global_var( y_var = layers.create_global_var(
shape=shape, value=y_val, dtype=dtype, persistable=True) name='y', shape=shape, value=0.0, dtype=dtype, persistable=True)
out = fn(x_var, y_var) out = fn(x_var, y_var)
exe = fluid.Executor(place) exe = fluid.Executor(place)
exe.run(fluid.default_startup_program()) exe.run(fluid.default_startup_program())
fluid_out = exe.run(fluid.default_main_program(), fluid_out = exe.run(fluid.default_main_program(),
feed=[], feed={'x': x_data,
'y': y_data},
fetch_list=[out]) fetch_list=[out])
np.testing.assert_array_equal(python_out, fluid_out[0]) np.testing.assert_array_equal(python_out, fluid_out[0])
def test_override(self): def test_override(self):
cpu_place = fluid.CPUPlace() # compare func to check
test_data = [(lambda _a, _b: _a == _b, 0.1, 1.1, cpu_place, 'float32'), compare_fns = [
(lambda _a, _b: _a == _b, 1.2, 1.1, cpu_place, 'float32'), lambda _a, _b: _a == _b,
(lambda _a, _b: _a < _b, 0.1, 1.1, cpu_place, 'float32'), lambda _a, _b: _a == _b,
(lambda _a, _b: _a < _b, 2.1, 1.1, cpu_place, 'float32'), lambda _a, _b: _a < _b,
(lambda _a, _b: _a <= _b, 0.1, 1.1, cpu_place, 'float32'), lambda _a, _b: _a < _b,
(lambda _a, _b: _a <= _b, 1.1, 1.1, cpu_place, 'float32'), lambda _a, _b: _a <= _b,
(lambda _a, _b: _a >= _b, 1.1, 1.1, cpu_place, 'float32')] lambda _a, _b: _a <= _b,
lambda _a, _b: _a >= _b,
main_program = framework.Program() ]
startup_program = framework.Program()
# places to check
with framework.program_guard(main_program, startup_program): places = [fluid.CPUPlace()]
for fn, x_val, y_val, place, dtype in test_data: if fluid.core.is_compiled_with_cuda():
self.check_result(fn, x_val, y_val, place, dtype) places.append(fluid.CUDAPlace(0))
# dtypes to check
dtypes = ['int32', 'float32']
for place in places:
for dtype in dtypes:
for compare_fn in compare_fns:
with framework.program_guard(framework.Program(),
gframework.Program()):
self.check_result(compare_fn, place, dtype)
if __name__ == '__main__': if __name__ == '__main__':
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册