Skip to content

P
Paddle

BaiXuePrincess / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 1
Fork 0
P
Paddle
未验证 提交 f3a93e51 编写于 作者: C ccrrong 提交者： GitHub
浏览文件
操作

Move conv3d from fluid to static.nn.common (#48266) 

* move conv3d

* remove unused import
上级 6138331d
隐藏空白更改
内联 并排
Showing with 370 addition and 367 deletion
python/paddle/fluid/layers/nn.py
浏览文件 @ f3a93e51
......@@ -69,7 +69,6 @@ __all__ = [
'crf_decoding',
'cos_sim',
'conv2d',
'conv3d',
'softmax',
'pool2d',
'pool3d',
......@@ -1683,314 +1682,6 @@ def conv2d(
return helper.append_activation(pre_act)
def conv3d(
input,
num_filters,
filter_size,
stride=1,
padding=0,
dilation=1,
groups=None,
param_attr=None,
bias_attr=None,
use_cudnn=True,
act=None,
name=None,
data_format="NCDHW",
):
r"""
:api_attr: Static Graph
The convolution3D layer calculates the output based on the input, filter
and strides, paddings, dilations, groups parameters. Input(Input) and
Output(Output) are in NCDHW or NDHWC format. Where N is batch size C is the number of
channels, D is the depth of the feature, H is the height of the feature,
and W is the width of the feature. Convlution3D is similar with Convlution2D
but adds one dimension(depth). If bias attribution and activation type are
provided, bias is added to the output of the convolution, and the
corresponding activation function is applied to the final result.
For each input :math:`X`, the equation is:
.. math::
Out = \sigma (W \\ast X + b)
In the above equation:
* :math:`X`: Input value, a tensor with NCDHW or NDHWC format.
* :math:`W`: Filter value, a tensor with MCDHW format.
* :math:`\\ast`: Convolution operation.
* :math:`b`: Bias value, a 2-D tensor with shape [M, 1].
* :math:`\\sigma`: Activation function.
* :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different.
Example:
- Input:
Input shape: :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})`
Filter shape: :math:`(C_{out}, C_{in}, D_f, H_f, W_f)`
- Output:
Output shape: :math:`(N, C_{out}, D_{out}, H_{out}, W_{out})`
Where
.. math::
D_{out}&= \\frac{(D_{in} + 2 * paddings[0] - (dilations[0] * (D_f - 1) + 1))}{strides[0]} + 1 \\\\
H_{out}&= \\frac{(H_{in} + 2 * paddings[1] - (dilations[1] * (H_f - 1) + 1))}{strides[1]} + 1 \\\\
W_{out}&= \\frac{(W_{in} + 2 * paddings[2] - (dilations[2] * (W_f - 1) + 1))}{strides[2]} + 1
Args:
input (Tensor): The input is 5-D Tensor with shape [N, C, D, H, W], the data
type of input is float16 or float32 or float64.
num_filters(int): The number of filter. It is as same as the output
image channel.
filter_size (int|tuple): The filter size. If filter_size is a tuple,
it must contain three integers, (filter_size_depth, filter_size_height,
filter_size_width). Otherwise, filter_size_depth = filter_size_height = \
filter_size_width = filter_size.
stride (int|tuple): The stride size. It means the stride in convolution. If stride is a
tuple, it must contain three integers, (stride_depth, stride_height, stride_width).
Otherwise, stride_depth = stride_height = stride_width = stride. Default: stride = 1.
padding (string|int|list|tuple): The padding size. It means the number of zero-paddings
on both sides for each dimension. If `padding` is a string, either 'VALID' or
'SAME' which is the padding algorithm. If padding size is a tuple or list,
it could be in three forms: `[pad_depth, pad_height, pad_width]` or
`[pad_depth_front, pad_depth_back, pad_height_top, pad_height_bottom, pad_width_left, pad_width_right]`,
and when `data_format` is `"NCDHW"`, `pool_padding` can be in the form
`[[0,0], [0,0], [pad_depth_front, pad_depth_back], [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right]]`.
when `data_format` is `"NDHWC"`, `pool_padding` can be in the form
`[[0,0], [pad_depth_front, pad_depth_back], [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right], [0,0]]`.
Default: padding = 0.
dilation (int|tuple): The dilation size. It means the spacing between the kernel points.
If dilation is a tuple, it must contain three integers, (dilation_depth, dilation_height,
dilation_width). Otherwise, dilation_depth = dilation_height = dilation_width = dilation.
Default: dilation = 1.
groups (int): The groups number of the Conv3d Layer. According to grouped
convolution in Alex Krizhevsky's Deep CNN paper: when group=2,
the first half of the filters is only connected to the first half
of the input channels, while the second half of the filters is only
connected to the second half of the input channels. Default: groups=1
param_attr (ParamAttr|None): The parameter attribute for learnable parameters/weights
of conv3d. If it is set to None or one attribute of ParamAttr, conv3d
will create ParamAttr as param_attr. If it is set to None, the parameter
is initialized with :math:`Normal(0.0, std)`, and the :math:`std` is
:math:`(\\frac{2.0 }{filter\_elem\_num})^{0.5}`. Default: None.
bias_attr (ParamAttr|bool|None): The parameter attribute for the bias of conv3d.
If it is set to False, no bias will be added to the output units.
If it is set to None or one attribute of ParamAttr, conv3d
will create ParamAttr as bias_attr. If the Initializer of the bias_attr
is not set, the bias is initialized zero. Default: None.
use_cudnn (bool): Use cudnn kernel or not, it is valid only when the cudnn
library is installed. Default: True
act (str): Activation type, if it is set to None, activation is not appended.
Default: None.
name(str|None): For detailed information, please refer
to :ref:`api_guide_Name`. Usually name is no need to set and
None by default.
data_format (str, optional): Specify the data format of the input, and the data format of the output
will be consistent with that of the input. An optional string from: `"NCHW"`, `"NHWC"`.
The default is `"NCHW"`. When it is `"NCHW"`, the data is stored in the order of:
`[batch_size, input_channels, input_height, input_width]`.
Returns:
A Variable holding Tensor representing the conv3d, whose data type is
the same with input. If act is None, the tensor variable storing the
convolution result, and if act is not None, the tensor variable storing
convolution and non-linearity activation result.
Raises:
ValueError: If the type of `use_cudnn` is not bool.
ValueError: If `data_format` is not "NCDHW" or "NDHWC".
ValueError: If the channel dimmention of the input is less than or equal to zero.
ValueError: If `padding` is a string, but not "SAME" or "VALID".
ValueError: If `padding` is a tuple, but the element corresponding to the input's batch size is not 0
or the element corresponding to the input's channel is not 0.
ShapeError: If the input is not 5-D Tensor.
ShapeError: If the input's dimension size and filter's dimension size not equal.
ShapeError: If the dimension size of input minus the size of `stride` is not 2.
ShapeError: If the number of input channels is not equal to filter's channels * groups.
ShapeError: If the number of output channels is not be divided by groups.
Examples:
.. code-block:: python
import paddle
import numpy as np
paddle.enable_static()
data = paddle.static.data(name='data', shape=[None, 3, 12, 32, 32], dtype='float32')
param_attr = paddle.framework.ParamAttr(name='conv3d.weight', initializer=paddle.nn.initializer.XavierNormal(), learning_rate=0.001)
res = paddle.static.nn.conv3d(input=data, num_filters=2, filter_size=3, act="relu", param_attr=param_attr)
place = paddle.CPUPlace()
exe = paddle.static.Executor(place)
exe.run(paddle.static.default_startup_program())
x = np.random.rand(1, 3, 12, 32, 32).astype("float32")
output = exe.run(feed={"data": x}, fetch_list=[res])
print(output)
"""
l_type = 'conv3d'
assert param_attr is not False, "param_attr should not be False here."
helper = LayerHelper(l_type, **locals())
dtype = helper.input_dtype()
if not isinstance(use_cudnn, bool):
raise ValueError(
"Attr(use_cudnn) should be True or False. Received "
"Attr(use_cudnn): %s. " % str(use_cudnn)
)
if data_format not in ["NCDHW", "NDHWC"]:
raise ValueError(
"Attr(data_format) should be 'NCDHW' or 'NDHWC'. Received "
"Attr(data_format): %s." % str(data_format)
)
channel_last = data_format == "NDHWC"
if len(input.shape) != 5:
raise ValueError(
"Input should be 5D tensor, but received input with the shape of {}".format(
input.shape
)
)
num_channels = input.shape[4] if channel_last else input.shape[1]
if num_channels < 0:
raise ValueError(
"The channel dimmention of the input(%s) should be defined. "
"Received: %s." % (str(input.shape), str(num_channels))
)
if groups is None:
num_filter_channels = num_channels
elif groups <= 0:
raise ValueError(
"the groups of conv3d should be greater than 0. Received groups: {}".format(
groups
)
)
else:
if num_channels % groups != 0:
raise ValueError(
"The number of input channels must be divisible by Attr(groups). "
"Received: number of channels(%s), groups(%s)."
% (str(num_channels), str(groups))
)
num_filter_channels = num_channels // groups
filter_size = utils.convert_to_list(filter_size, 3, 'filter_size')
stride = utils.convert_to_list(stride, 3, 'stride')
dilation = utils.convert_to_list(dilation, 3, 'dilation')
def _update_padding(padding, data_format):
def is_list_or_tuple(ele):
if isinstance(ele, list) or isinstance(ele, tuple):
return True
return False
if is_list_or_tuple(padding) and len(padding) == 5:
if is_list_or_tuple(padding[0]) and (data_format == "NCDHW"):
if not (padding[0] == [0, 0] and padding[1] == [0, 0]):
raise ValueError(
"Non-zero padding(%s) in the batch or channel dimensions "
"is not supported." % str(padding)
)
padding = padding[2:5]
padding = [ele for a_list in padding for ele in a_list]
elif is_list_or_tuple(padding[0]) and (data_format == "NDHWC"):
if not (padding[0] == [0, 0] and padding[4] == [0, 0]):
raise ValueError(
"Non-zero padding(%s) in the batch or channel dimensions "
"is not supported." % str(padding)
)
padding = padding[1:4]
padding = [ele for a_list in padding for ele in a_list]
padding = utils.convert_to_list(padding, 6, 'padding')
if utils._is_symmetric_padding(padding, 3):
padding = [padding[0], padding[2], padding[4]]
elif is_list_or_tuple(padding) and len(padding) == 6:
padding = utils.convert_to_list(padding, 6, 'padding')
if utils._is_symmetric_padding(padding, 3):
padding = [padding[0], padding[2], padding[4]]
else:
padding = utils.convert_to_list(padding, 3, 'padding')
return padding
padding_algorithm = "EXPLICIT"
if isinstance(padding, str):
padding = padding.upper()
if padding not in ["SAME", "VALID"]:
raise ValueError(
"Unknown padding: '%s'. It can only be 'SAME' or 'VALID'."
% str(padding)
)
if padding == "VALID":
padding_algorithm = "VALID"
padding = [0, 0, 0]
elif padding == "SAME":
padding_algorithm = "SAME"
padding = [0, 0, 0]
padding = _update_padding(padding, data_format)
input_shape = input.shape
filter_shape = [num_filters, num_filter_channels] + filter_size
def _get_default_param_initializer():
filter_elem_num = (
filter_size[0] * filter_size[1] * filter_size[2] * num_channels
)
if filter_elem_num <= 0:
raise ValueError(
"Invalid filter number, excepted number is larger than 0, but"
" received {}, please check the input shape and "
"filter size.".format(filter_elem_num)
)
std = (2.0 / filter_elem_num) ** 0.5
return Normal(0.0, std, 0)
filter_param = helper.create_parameter(
attr=helper.param_attr,
shape=filter_shape,
dtype=dtype,
default_initializer=_get_default_param_initializer(),
)
pre_bias = helper.create_variable_for_type_inference(dtype)
helper.append_op(
type=l_type,
inputs={
'Input': input,
'Filter': filter_param,
},
outputs={"Output": pre_bias},
attrs={
'strides': stride,
'paddings': padding,
'dilations': dilation,
'groups': groups,
'use_cudnn': use_cudnn,
'use_mkldnn': False,
"padding_algorithm": padding_algorithm,
"data_format": data_format,
},
)
if data_format == 'NCDHW':
pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2)
else:
pre_act = helper.append_bias_op(pre_bias, dim_start=4, dim_end=5)
return helper.append_activation(pre_act)
@templatedoc()
def pool2d(
input,
......
python/paddle/fluid/tests/unittests/ir/inference/test_trt_conv3d_op.py
浏览文件 @ f3a93e51
......@@ -15,6 +15,7 @@
import unittest
import numpy as np
from inference_pass_test import InferencePassTest
import paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
from paddle.fluid.core import PassVersionChecker
......@@ -29,7 +30,7 @@ class TensorRTSubgraphPassConv3dTest(InferencePassTest):
data = fluid.data(
name="data", shape=[-1, 3, 6, 32, 32], dtype="float32"
)
conv_out = fluid.layers.conv3d(
conv_out = paddle.static.nn.conv3d(
input=data,
num_filters=self.conv_num_filters,
filter_size=self.conv_filter_size,
......@@ -113,7 +114,7 @@ class DynamicShapeTensorRTSubgraphPassConv3dTest(InferencePassTest):
data = fluid.data(
name="data", shape=[-1, 6, -1, -1, -1], dtype="float32"
)
conv_out = fluid.layers.conv3d(
conv_out = paddle.static.nn.conv3d(
input=data,
num_filters=self.conv_num_filters,
filter_size=self.conv_filter_size,
......
python/paddle/fluid/tests/unittests/npu/test_conv3d_op_npu.py
浏览文件 @ f3a93e51
......@@ -364,7 +364,7 @@ class TestConv3DAPI(unittest.TestCase):
dtype="float32",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NDHWC,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -375,7 +375,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -386,7 +386,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -397,7 +397,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NDHWC,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -408,7 +408,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NDHWC",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -419,7 +419,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -442,7 +442,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: cudnn
def run_1():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -458,7 +458,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: data_format
def run_2():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -474,7 +474,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: padding
def run_3():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -489,7 +489,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
self.assertRaises(ValueError, run_3)
def run_4():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -504,7 +504,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
self.assertRaises(ValueError, run_4)
def run_5():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=0,
......@@ -527,7 +527,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
)
def run_6():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=x,
num_filters=3,
filter_size=3,
......@@ -543,7 +543,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: groups
def run_7():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -559,7 +559,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: filter num
def run_8():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=0,
filter_size=0,
......
python/paddle/fluid/tests/unittests/test_conv3d_layer.py
浏览文件 @ f3a93e51
......@@ -101,7 +101,7 @@ class Conv3DTestCase(unittest.TestCase):
bias_attr = False
else:
bias_attr = I.NumpyArrayInitializer(self.bias)
y_var = fluid.layers.conv3d(
y_var = paddle.static.nn.conv3d(
x_var,
self.num_filters,
self.filter_size,
......
python/paddle/fluid/tests/unittests/test_conv3d_op.py
浏览文件 @ f3a93e51
......@@ -878,7 +878,7 @@ class TestConv3DAPI(unittest.TestCase):
dtype="float32",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NDHWC,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -889,7 +889,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -900,7 +900,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -911,7 +911,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NDHWC,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -922,7 +922,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NDHWC",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -933,7 +933,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -956,7 +956,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: cudnn
def run_1():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -972,7 +972,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: data_format
def run_2():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -988,7 +988,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: padding
def run_3():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -1003,7 +1003,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
self.assertRaises(ValueError, run_3)
def run_4():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -1018,7 +1018,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
self.assertRaises(ValueError, run_4)
def run_5():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=0,
......@@ -1041,7 +1041,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
)
def run_6():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=x,
num_filters=3,
filter_size=3,
......@@ -1057,7 +1057,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: groups
def run_7():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -1073,7 +1073,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: filter num
def run_8():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=0,
filter_size=0,
......
python/paddle/fluid/tests/unittests/test_conv_nn_grad.py
浏览文件 @ f3a93e51
......@@ -110,7 +110,7 @@ class TestConv3DDoubleGradCheck(unittest.TestCase):
eps = 0.005
dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64
x = layers.data('x', shape, False, dtype)
y = layers.conv3d(x, 2, 1, bias_attr=False)
y = paddle.static.nn.conv3d(x, 2, 1, bias_attr=False)
x_arr = np.random.uniform(-1, 1, shape).astype(dtype)
w = fluid.default_main_program().global_block().all_parameters()
......@@ -137,7 +137,7 @@ class TestConv3DDoubleGradCheckTest1(unittest.TestCase):
eps = 0.005
dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64
x = layers.data('x', shape, False, dtype)
y = layers.conv3d(x, 2, 1, padding=1, bias_attr=False)
y = paddle.static.nn.conv3d(x, 2, 1, padding=1, bias_attr=False)
x_arr = np.random.uniform(-1, 1, shape).astype(dtype)
w = fluid.default_main_program().global_block().all_parameters()
......@@ -332,7 +332,7 @@ class TestConv3DDoubleGradCheck_AsyPadding(unittest.TestCase):
eps = 0.005
dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64
x = layers.data('x', shape, False, dtype)
y = layers.conv3d(
y = paddle.static.nn.conv3d(
input=x,
num_filters=2,
filter_size=1,
......@@ -365,7 +365,7 @@ class TestConv3DoubleGradCheck_PaddingSAME(unittest.TestCase):
eps = 0.005
dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64
x = layers.data('x', shape, False, dtype)
y = layers.conv3d(
y = paddle.static.nn.conv3d(
input=x,
num_filters=2,
filter_size=1,
......@@ -399,7 +399,7 @@ class TestConv3DoubleGradCheck_PaddingVALID(unittest.TestCase):
eps = 0.005
dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64
x = layers.data('x', shape, False, dtype)
y = layers.conv3d(
y = paddle.static.nn.conv3d(
input=x,
num_filters=2,
filter_size=1,
......@@ -432,7 +432,7 @@ class TestConv3DDoubleGradCheck_ChannelLast(unittest.TestCase):
eps = 0.005
dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64
x = layers.data('x', shape, False, dtype)
y = layers.conv3d(
y = paddle.static.nn.conv3d(
input=x,
num_filters=2,
filter_size=1,
......@@ -467,7 +467,7 @@ class TestConv3DDoubleGradCheck_ChannelLast_AsyPadding(unittest.TestCase):
eps = 0.005
dtype = np.float32 if fluid.core.is_compiled_with_rocm() else np.float64
x = layers.data('x', shape, False, dtype)
y = layers.conv3d(
y = paddle.static.nn.conv3d(
input=x,
num_filters=2,
filter_size=1,
......
python/paddle/fluid/tests/unittests/test_functional_conv3d.py
浏览文件 @ f3a93e51
......@@ -87,7 +87,7 @@ class TestFunctionalConv3D(TestCase):
(-1, self.in_channels, -1, -1, -1),
dtype=self.dtype,
)
y = fluid.layers.conv3d(
y = paddle.static.nn.conv3d(
x,
self.out_channels,
self.filter_shape,
......@@ -480,7 +480,7 @@ class TestFunctionalConv3DErrorCase11(TestCase):
with fluid.unique_name.guard():
with fluid.program_guard(main, start):
x = fluid.data("input", self.input.shape, dtype=paddle.float32)
y = fluid.layers.conv3d(
y = paddle.static.nn.conv3d(
x,
self.num_filters,
self.filter_size,
......
python/paddle/fluid/tests/unittests/test_imperative_load_static_param.py
浏览文件 @ f3a93e51
......@@ -51,10 +51,10 @@ class TestDygraphLoadStatic(unittest.TestCase):
conv3d_in = fluid.data(
name='conv3d_in', shape=[None, 3, 12, 32, 32], dtype='float32'
)
conv3d_out_1 = fluid.layers.conv3d(
conv3d_out_1 = paddle.static.nn.conv3d(
input=conv3d_in, num_filters=2, filter_size=3, act="relu"
)
conv3d_out_2 = fluid.layers.conv3d(
conv3d_out_2 = paddle.static.nn.conv3d(
input=conv3d_in, num_filters=2, filter_size=3, act="relu"
)
......
python/paddle/fluid/tests/unittests/test_layers.py
浏览文件 @ f3a93e51
......@@ -1688,7 +1688,9 @@ class TestLayer(LayerTest):
images = layers.data(
name='pixel', shape=[3, 6, 6, 6], dtype='float32'
)
ret = layers.conv3d(input=images, num_filters=3, filter_size=2)
ret = paddle.static.nn.conv3d(
input=images, num_filters=3, filter_size=2
)
static_ret = self.get_static_graph_result(
feed={'pixel': np.ones([2, 3, 6, 6, 6], dtype='float32')},
fetch_list=[ret],
......
python/paddle/fluid/tests/unittests/xpu/test_conv3d_op_xpu.py
浏览文件 @ f3a93e51
......@@ -525,7 +525,7 @@ class TestConv3DAPI(unittest.TestCase):
dtype="float32",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NDHWC,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -536,7 +536,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -547,7 +547,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -558,7 +558,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NDHWC,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -569,7 +569,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NDHWC",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -580,7 +580,7 @@ class TestConv3DAPI(unittest.TestCase):
data_format="NCDHW",
)
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input_NCDHW,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -603,7 +603,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: cudnn
def run_1():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -619,7 +619,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: data_format
def run_2():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=[3, 3, 3],
......@@ -635,7 +635,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: padding
def run_3():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -650,7 +650,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
self.assertRaises(ValueError, run_3)
def run_4():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -665,7 +665,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
self.assertRaises(ValueError, run_4)
def run_5():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=0,
......@@ -688,7 +688,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
)
def run_6():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=x,
num_filters=3,
filter_size=3,
......@@ -704,7 +704,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: groups
def run_7():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=3,
filter_size=3,
......@@ -720,7 +720,7 @@ class TestConv3DAPI_Error(unittest.TestCase):
# ValueError: filter num
def run_8():
fluid.layers.conv3d(
paddle.static.nn.conv3d(
input=input,
num_filters=0,
filter_size=0,
......
python/paddle/static/nn/__init__.py
浏览文件 @ f3a93e51
......@@ -14,6 +14,7 @@
from .common import fc # noqa: F401
from .common import deform_conv2d # noqa: F401
from .common import conv3d # noqa: F401
from .common import conv2d_transpose # noqa: F401
from .common import conv3d_transpose # noqa: F401
......@@ -22,7 +23,6 @@ from ...fluid.layers import bilinear_tensor_product # noqa: F401
from ...fluid.layers import case # noqa: F401
from ...fluid.layers import cond # noqa: F401
from ...fluid.layers import conv2d # noqa: F401
from ...fluid.layers import conv3d # noqa: F401
from ...fluid.layers import create_parameter # noqa: F401
from ...fluid.layers import crf_decoding # noqa: F401
from ...fluid.layers import data_norm # noqa: F401
......
python/paddle/static/nn/common.py
浏览文件 @ f3a93e51
......@@ -13,6 +13,7 @@
# limitations under the License.
import paddle
from paddle.fluid.initializer import Normal
from paddle.fluid.framework import static_only, Variable, _non_static_mode
from paddle.fluid.data_feeder import check_dtype
......@@ -176,6 +177,314 @@ def fc(
)
def conv3d(
input,
num_filters,
filter_size,
stride=1,
padding=0,
dilation=1,
groups=None,
param_attr=None,
bias_attr=None,
use_cudnn=True,
act=None,
name=None,
data_format="NCDHW",
):
r"""
:api_attr: Static Graph
The convolution3D layer calculates the output based on the input, filter
and strides, paddings, dilations, groups parameters. Input(Input) and
Output(Output) are in NCDHW or NDHWC format. Where N is batch size C is the number of
channels, D is the depth of the feature, H is the height of the feature,
and W is the width of the feature. Convlution3D is similar with Convlution2D
but adds one dimension(depth). If bias attribution and activation type are
provided, bias is added to the output of the convolution, and the
corresponding activation function is applied to the final result.
For each input :math:`X`, the equation is:
.. math::
Out = \sigma (W \\ast X + b)
In the above equation:
* :math:`X`: Input value, a tensor with NCDHW or NDHWC format.
* :math:`W`: Filter value, a tensor with MCDHW format.
* :math:`\\ast`: Convolution operation.
* :math:`b`: Bias value, a 2-D tensor with shape [M, 1].
* :math:`\\sigma`: Activation function.
* :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different.
Example:
- Input:
Input shape: :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})`
Filter shape: :math:`(C_{out}, C_{in}, D_f, H_f, W_f)`
- Output:
Output shape: :math:`(N, C_{out}, D_{out}, H_{out}, W_{out})`
Where
.. math::
D_{out}&= \\frac{(D_{in} + 2 * paddings[0] - (dilations[0] * (D_f - 1) + 1))}{strides[0]} + 1 \\\\
H_{out}&= \\frac{(H_{in} + 2 * paddings[1] - (dilations[1] * (H_f - 1) + 1))}{strides[1]} + 1 \\\\
W_{out}&= \\frac{(W_{in} + 2 * paddings[2] - (dilations[2] * (W_f - 1) + 1))}{strides[2]} + 1
Args:
input (Tensor): The input is 5-D Tensor with shape [N, C, D, H, W], the data
type of input is float16 or float32 or float64.
num_filters(int): The number of filter. It is as same as the output
image channel.
filter_size (int|tuple): The filter size. If filter_size is a tuple,
it must contain three integers, (filter_size_depth, filter_size_height,
filter_size_width). Otherwise, filter_size_depth = filter_size_height = \
filter_size_width = filter_size.
stride (int|tuple): The stride size. It means the stride in convolution. If stride is a
tuple, it must contain three integers, (stride_depth, stride_height, stride_width).
Otherwise, stride_depth = stride_height = stride_width = stride. Default: stride = 1.
padding (string|int|list|tuple): The padding size. It means the number of zero-paddings
on both sides for each dimension. If `padding` is a string, either 'VALID' or
'SAME' which is the padding algorithm. If padding size is a tuple or list,
it could be in three forms: `[pad_depth, pad_height, pad_width]` or
`[pad_depth_front, pad_depth_back, pad_height_top, pad_height_bottom, pad_width_left, pad_width_right]`,
and when `data_format` is `"NCDHW"`, `pool_padding` can be in the form
`[[0,0], [0,0], [pad_depth_front, pad_depth_back], [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right]]`.
when `data_format` is `"NDHWC"`, `pool_padding` can be in the form
`[[0,0], [pad_depth_front, pad_depth_back], [pad_height_top, pad_height_bottom], [pad_width_left, pad_width_right], [0,0]]`.
Default: padding = 0.
dilation (int|tuple): The dilation size. It means the spacing between the kernel points.
If dilation is a tuple, it must contain three integers, (dilation_depth, dilation_height,
dilation_width). Otherwise, dilation_depth = dilation_height = dilation_width = dilation.
Default: dilation = 1.
groups (int): The groups number of the Conv3d Layer. According to grouped
convolution in Alex Krizhevsky's Deep CNN paper: when group=2,
the first half of the filters is only connected to the first half
of the input channels, while the second half of the filters is only
connected to the second half of the input channels. Default: groups=1
param_attr (ParamAttr|None): The parameter attribute for learnable parameters/weights
of conv3d. If it is set to None or one attribute of ParamAttr, conv3d
will create ParamAttr as param_attr. If it is set to None, the parameter
is initialized with :math:`Normal(0.0, std)`, and the :math:`std` is
:math:`(\\frac{2.0 }{filter\_elem\_num})^{0.5}`. Default: None.
bias_attr (ParamAttr|bool|None): The parameter attribute for the bias of conv3d.
If it is set to False, no bias will be added to the output units.
If it is set to None or one attribute of ParamAttr, conv3d
will create ParamAttr as bias_attr. If the Initializer of the bias_attr
is not set, the bias is initialized zero. Default: None.
use_cudnn (bool): Use cudnn kernel or not, it is valid only when the cudnn
library is installed. Default: True
act (str): Activation type, if it is set to None, activation is not appended.
Default: None.
name(str|None): For detailed information, please refer
to :ref:`api_guide_Name`. Usually name is no need to set and
None by default.
data_format (str, optional): Specify the data format of the input, and the data format of the output
will be consistent with that of the input. An optional string from: `"NCHW"`, `"NHWC"`.
The default is `"NCHW"`. When it is `"NCHW"`, the data is stored in the order of:
`[batch_size, input_channels, input_height, input_width]`.
Returns:
A Variable holding Tensor representing the conv3d, whose data type is
the same with input. If act is None, the tensor variable storing the
convolution result, and if act is not None, the tensor variable storing
convolution and non-linearity activation result.
Raises:
ValueError: If the type of `use_cudnn` is not bool.
ValueError: If `data_format` is not "NCDHW" or "NDHWC".
ValueError: If the channel dimmention of the input is less than or equal to zero.
ValueError: If `padding` is a string, but not "SAME" or "VALID".
ValueError: If `padding` is a tuple, but the element corresponding to the input's batch size is not 0
or the element corresponding to the input's channel is not 0.
ShapeError: If the input is not 5-D Tensor.
ShapeError: If the input's dimension size and filter's dimension size not equal.
ShapeError: If the dimension size of input minus the size of `stride` is not 2.
ShapeError: If the number of input channels is not equal to filter's channels * groups.
ShapeError: If the number of output channels is not be divided by groups.
Examples:
.. code-block:: python
import paddle
import numpy as np
paddle.enable_static()
data = paddle.static.data(name='data', shape=[None, 3, 12, 32, 32], dtype='float32')
param_attr = paddle.framework.ParamAttr(name='conv3d.weight', initializer=paddle.nn.initializer.XavierNormal(), learning_rate=0.001)
res = paddle.static.nn.conv3d(input=data, num_filters=2, filter_size=3, act="relu", param_attr=param_attr)
place = paddle.CPUPlace()
exe = paddle.static.Executor(place)
exe.run(paddle.static.default_startup_program())
x = np.random.rand(1, 3, 12, 32, 32).astype("float32")
output = exe.run(feed={"data": x}, fetch_list=[res])
print(output)
"""
l_type = 'conv3d'
assert param_attr is not False, "param_attr should not be False here."
helper = LayerHelper(l_type, **locals())
dtype = helper.input_dtype()
if not isinstance(use_cudnn, bool):
raise ValueError(
"Attr(use_cudnn) should be True or False. Received "
"Attr(use_cudnn): %s. " % str(use_cudnn)
)
if data_format not in ["NCDHW", "NDHWC"]:
raise ValueError(
"Attr(data_format) should be 'NCDHW' or 'NDHWC'. Received "
"Attr(data_format): %s." % str(data_format)
)
channel_last = data_format == "NDHWC"
if len(input.shape) != 5:
raise ValueError(
"Input should be 5D tensor, but received input with the shape of {}".format(
input.shape
)
)
num_channels = input.shape[4] if channel_last else input.shape[1]
if num_channels < 0:
raise ValueError(
"The channel dimmention of the input(%s) should be defined. "
"Received: %s." % (str(input.shape), str(num_channels))
)
if groups is None:
num_filter_channels = num_channels
elif groups <= 0:
raise ValueError(
"the groups of conv3d should be greater than 0. Received groups: {}".format(
groups
)
)
else:
if num_channels % groups != 0:
raise ValueError(
"The number of input channels must be divisible by Attr(groups). "
"Received: number of channels(%s), groups(%s)."
% (str(num_channels), str(groups))
)
num_filter_channels = num_channels // groups
filter_size = utils.convert_to_list(filter_size, 3, 'filter_size')
stride = utils.convert_to_list(stride, 3, 'stride')
dilation = utils.convert_to_list(dilation, 3, 'dilation')
def _update_padding(padding, data_format):
def is_list_or_tuple(ele):
if isinstance(ele, list) or isinstance(ele, tuple):
return True
return False
if is_list_or_tuple(padding) and len(padding) == 5:
if is_list_or_tuple(padding[0]) and (data_format == "NCDHW"):
if not (padding[0] == [0, 0] and padding[1] == [0, 0]):
raise ValueError(
"Non-zero padding(%s) in the batch or channel dimensions "
"is not supported." % str(padding)
)
padding = padding[2:5]
padding = [ele for a_list in padding for ele in a_list]
elif is_list_or_tuple(padding[0]) and (data_format == "NDHWC"):
if not (padding[0] == [0, 0] and padding[4] == [0, 0]):
raise ValueError(
"Non-zero padding(%s) in the batch or channel dimensions "
"is not supported." % str(padding)
)
padding = padding[1:4]
padding = [ele for a_list in padding for ele in a_list]
padding = utils.convert_to_list(padding, 6, 'padding')
if utils._is_symmetric_padding(padding, 3):
padding = [padding[0], padding[2], padding[4]]
elif is_list_or_tuple(padding) and len(padding) == 6:
padding = utils.convert_to_list(padding, 6, 'padding')
if utils._is_symmetric_padding(padding, 3):
padding = [padding[0], padding[2], padding[4]]
else:
padding = utils.convert_to_list(padding, 3, 'padding')
return padding
padding_algorithm = "EXPLICIT"
if isinstance(padding, str):
padding = padding.upper()
if padding not in ["SAME", "VALID"]:
raise ValueError(
"Unknown padding: '%s'. It can only be 'SAME' or 'VALID'."
% str(padding)
)
if padding == "VALID":
padding_algorithm = "VALID"
padding = [0, 0, 0]
elif padding == "SAME":
padding_algorithm = "SAME"
padding = [0, 0, 0]
padding = _update_padding(padding, data_format)
input_shape = input.shape
filter_shape = [num_filters, num_filter_channels] + filter_size
def _get_default_param_initializer():
filter_elem_num = (
filter_size[0] * filter_size[1] * filter_size[2] * num_channels
)
if filter_elem_num <= 0:
raise ValueError(
"Invalid filter number, excepted number is larger than 0, but"
" received {}, please check the input shape and "
"filter size.".format(filter_elem_num)
)
std = (2.0 / filter_elem_num) ** 0.5
return Normal(0.0, std, 0)
filter_param = helper.create_parameter(
attr=helper.param_attr,
shape=filter_shape,
dtype=dtype,
default_initializer=_get_default_param_initializer(),
)
pre_bias = helper.create_variable_for_type_inference(dtype)
helper.append_op(
type=l_type,
inputs={
'Input': input,
'Filter': filter_param,
},
outputs={"Output": pre_bias},
attrs={
'strides': stride,
'paddings': padding,
'dilations': dilation,
'groups': groups,
'use_cudnn': use_cudnn,
'use_mkldnn': False,
"padding_algorithm": padding_algorithm,
"data_format": data_format,
},
)
if data_format == 'NCDHW':
pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2)
else:
pre_act = helper.append_bias_op(pre_bias, dim_start=4, dim_end=5)
return helper.append_activation(pre_act)
def conv2d_transpose(
input,
num_filters,
......
Markdown is supported
0% 或 .
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 或