Skip to content

Mutiply bug allow non-tensor data input #27690

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zhhsplendid merged 4 commits into from
Oct 12, 2020
Merged

Mutiply bug allow non-tensor data input #27690

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
d3022bb
add implicitly paddle.tensor transform
Joejiong Sep 28, 2020
28aa96f
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…
Joejiong Sep 28, 2020
8f6f236
add static implicit tranformation
Joejiong Sep 29, 2020
ed1a0e4
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…
Joejiong Sep 29, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
75 changes: 75 additions & 0 deletions python/paddle/fluid/tests/unittests/test_multiply.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,7 @@ class TestMultiplyAPI(unittest.TestCase):

def __run_static_graph_case(self, x_data, y_data, axis=-1):
with program_guard(Program(), Program()):
paddle.enable_static()
Copy link
Member

@zhhsplendid zhhsplendid Oct 12, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

program是静态图才有的，所以最好paddle.enable_static写在with program_guard前面

x = paddle.static.data(
name='x', shape=x_data.shape, dtype=x_data.dtype)
y = paddle.static.data(
Expand All @@ -42,66 +43,140 @@ def __run_static_graph_case(self, x_data, y_data, axis=-1):
res = outs[0]
return res

def __run_static_graph_case_with_numpy_input(self, x_data, y_data, axis=-1):
with program_guard(Program(), Program()):
Copy link
Member

@zhhsplendid zhhsplendid Oct 12, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

program是静态图才有的，所以最好paddle.enable_static写在with program_guard前面

paddle.enable_static()

res = tensor.multiply(x_data, y_data, axis=axis)
place = fluid.CUDAPlace(0) if fluid.core.is_compiled_with_cuda(
) else fluid.CPUPlace()
exe = fluid.Executor(place)
Copy link
Member

@zhhsplendid zhhsplendid Oct 12, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

将来可以尽量用2.0 迁移的写法，比如paddle.CPUPlace, paddle.CUDAPlace，paddle.static.Executor

outs = exe.run(fluid.default_main_program(),
feed={'x': x_data,
'y': y_data},
fetch_list=[res])
res = outs[0]
return res

def __run_dynamic_graph_case(self, x_data, y_data, axis=-1):
paddle.disable_static()
x = paddle.to_tensor(x_data)
y = paddle.to_tensor(y_data)
res = paddle.multiply(x, y, axis=axis)
return res.numpy()

def __run_dynamic_graph_case_with_numpy_input(self, x_data, y_data,
axis=-1):
paddle.disable_static()
res = paddle.multiply(x_data, y_data, axis=axis)
return res.numpy()

def test_multiply(self):
"""test_multiply."""
np.random.seed(7)

# test static computation graph: 1-d array
x_data = np.random.rand(200)
y_data = np.random.rand(200)
res = self.__run_static_graph_case(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test static computation graph: 1-d array
Copy link
Member

@zhhsplendid zhhsplendid Oct 12, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

像我以前说过，python里面用注释分隔测试并不是最好的方法，你可以直接创建一个测试method:

def test_static_multiply_1d(self) 这类的

x_data = np.random.rand(200)
y_data = np.random.rand(200)
res = self.__run_static_graph_case_with_numpy_input(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test static computation graph: 2-d array
x_data = np.random.rand(2, 500)
y_data = np.random.rand(2, 500)
res = self.__run_static_graph_case(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test static computation graph with_primitives: 2-d array
x_data = np.random.rand(2, 500)
y_data = np.random.rand(2, 500)
res = self.__run_static_graph_case_with_numpy_input(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test static computation graph: broadcast
x_data = np.random.rand(2, 500)
y_data = np.random.rand(500)
res = self.__run_static_graph_case(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test static computation graph with_primitives: broadcast
x_data = np.random.rand(2, 500)
y_data = np.random.rand(500)
res = self.__run_static_graph_case_with_numpy_input(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test static computation graph: broadcast with axis
x_data = np.random.rand(2, 300, 40)
y_data = np.random.rand(300)
res = self.__run_static_graph_case(x_data, y_data, axis=1)
expected = np.multiply(x_data, y_data[..., np.newaxis])
self.assertTrue(np.allclose(res, expected))

# test static computation graph with_primitives: broadcast with axis
x_data = np.random.rand(2, 300, 40)
y_data = np.random.rand(300)
res = self.__run_static_graph_case_with_numpy_input(
x_data, y_data, axis=1)
expected = np.multiply(x_data, y_data[..., np.newaxis])
self.assertTrue(np.allclose(res, expected))

# test dynamic computation graph: 1-d array
x_data = np.random.rand(200)
y_data = np.random.rand(200)
res = self.__run_dynamic_graph_case(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test dynamic numpy input computation graph: 1-d array
x_data = np.random.rand(200)
y_data = np.random.rand(200)
res = self.__run_dynamic_graph_case_with_numpy_input(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test dynamic computation graph: 2-d array
x_data = np.random.rand(20, 50)
y_data = np.random.rand(20, 50)
res = self.__run_dynamic_graph_case(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test dynamic numpy input computation graph: 1-d array
x_data = np.random.rand(20, 50)
y_data = np.random.rand(20, 50)
res = self.__run_dynamic_graph_case_with_numpy_input(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test dynamic computation graph: broadcast
x_data = np.random.rand(2, 500)
y_data = np.random.rand(500)
res = self.__run_dynamic_graph_case(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test dynamic computation graph with numpy tensor: broadcast
x_data = np.random.rand(2, 500)
y_data = np.random.rand(500)
res = self.__run_dynamic_graph_case_with_numpy_input(x_data, y_data)
self.assertTrue(np.allclose(res, np.multiply(x_data, y_data)))

# test dynamic computation graph: broadcast with axis
x_data = np.random.rand(2, 300, 40)
y_data = np.random.rand(300)
res = self.__run_dynamic_graph_case(x_data, y_data, axis=1)
expected = np.multiply(x_data, y_data[..., np.newaxis])
self.assertTrue(np.allclose(res, expected))

# test dynamic computation graph with numpy tensor: broadcast with axis
x_data = np.random.rand(2, 300, 40)
y_data = np.random.rand(300)
res = self.__run_dynamic_graph_case_with_numpy_input(
x_data, y_data, axis=1)
expected = np.multiply(x_data, y_data[..., np.newaxis])
self.assertTrue(np.allclose(res, expected))


class TestMultiplyError(unittest.TestCase):
"""TestMultiplyError."""
Expand Down
12 changes: 12 additions & 0 deletions python/paddle/tensor/math.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -472,15 +472,27 @@ def multiply(x, y, axis=-1, name=None):
"""
op_type = 'elementwise_mul'
act = None

if x.dtype != y.dtype:
raise TypeError(
'Input tensors must be same type, but received type of x: %s, type of y: %s '
% (x.dtype, y.dtype))

if in_dygraph_mode():
if not isinstance(x, (paddle.Tensor)):
x = paddle.to_tensor(x)
if not isinstance(y, (paddle.Tensor)):
y = paddle.to_tensor(y)
return _elementwise_op_in_dygraph(
x, y, axis=axis, act=act, op_name=op_type)

if not isinstance(x, (paddle.Tensor, Variable)):
x = paddle.static.data(
name='x', shape=x.shape, dtype=x.dtype)
if not isinstance(y, (paddle.Tensor, Variable)):
y = paddle.static.data(
name='y', shape=y.shape, dtype=y.dtype)

return _elementwise_op(LayerHelper(op_type, **locals()))

def maximum(x, y, axis=-1, name=None):
Expand Down