diff --git a/python/paddle/trainer/config_parser.py b/python/paddle/trainer/config_parser.py
index 49b3c430e7..c0843a7357 100644
--- a/python/paddle/trainer/config_parser.py
+++ b/python/paddle/trainer/config_parser.py
@@ -901,20 +901,14 @@ class Conv3D(Cfg):
padding_z=None,
stride_z=None):
self.add_keys(locals())
- if filter_size_y is None:
- self.filter_size_y = filter_size
- if padding_y is None:
- self.padding_y = padding
- if stride_y is None:
- self.stride_y = stride
+ self.filter_size_y = filter_size_y if filter_size_y else filter_size
+ self.filter_size_z = filter_size_z if filter_size_z else filter_size
+ self.padding_y = padding_y if padding_y else padding
+ self.padding_z = padding_z if padding_z else padding
+ self.stride_y = stride_y if stride_y else stride
+ self.stride_z = stride_z if stride_z else stride
if output_x is not None:
config_assert(output_x <= 0)
- if filter_size_z is None:
- self.filter_size_z = filter_size
- if padding_z is None:
- self.padding_z = padding
- if stride_z is None:
- self.stride_z = stride
@config_class
@@ -1206,10 +1200,10 @@ def get_img_size(input_layer_name, channels):
def get_img3d_size(input_layer_name, channels):
input = g_layer_map[input_layer_name]
img_pixels = input.size / channels
- img_size = input.width if input.width > 0 else int(img_pixels**0.5)
- img_size_y = input.height if input.height > 0 else int(img_pixels /
- img_size)
- img_size_z = input.depth if input.depth > 1 else 1
+ img_size = input.width
+ img_size_y = input.height
+ img_size_z = input.depth
+
config_assert(
img_size * img_size_y * img_size_z == img_pixels,
"Input layer %s: Incorrect input image size %d * %d * %d for input image pixels %d"
@@ -2000,8 +1994,10 @@ class ConvLayer(ConvLayerBase):
layer_type = 'cudnn_conv'
-@config_layer('conv_3d')
-class Conv3DLayerBase(LayerBase):
+@config_layer('convt')
+class ConvTransLayerBase(LayerBase):
+ layer_type = 'convt'
+
def __init__(self,
name,
inputs=[],
@@ -2009,7 +2005,7 @@ class Conv3DLayerBase(LayerBase):
num_filters=None,
shared_biases=False,
**xargs):
- super(Conv3DLayerBase, self).__init__(
+ super(ConvTransLayerBase, self).__init__(
name, self.layer_type, 0, inputs=inputs, **xargs)
if num_filters is not None:
@@ -2018,12 +2014,17 @@ class Conv3DLayerBase(LayerBase):
use_gpu = int(g_command_config_args.get("use_gpu", 0))
parallel_nn = int(g_command_config_args.get("parallel_nn", 0))
- # Automatically select cudnn_type for GPU and exconv for CPU
- # if set type=conv, but still reserve the way user specify
- # exconv or cudnn_conv manually.
- if self.layer_type == "cudnn_conv3d":
- config_assert(use_gpu, "cudnn_conv3d only support GPU")
+ # Automatically select cudnn_type for GPU and exconvt for CPU
+ # if set type=exconvt, but still reserve the way user specify
+ # exconvt or cudnn_convt manually.
+ if self.layer_type == "cudnn_convt":
+ config_assert(use_gpu, "cudnn_convt only support GPU")
+ if (use_gpu == 1 and self.layer_type != "exconvt" and
+ (parallel_nn == 0 or self.config.device > -1)):
+ self.layer_type = "cudnn_convt"
+ else:
+ self.layer_type = "exconvt"
# need to specify layer in config
self.config.type = self.layer_type
@@ -2032,15 +2033,17 @@ class Conv3DLayerBase(LayerBase):
for input_index in xrange(len(self.inputs)):
input_layer = self.get_input_layer(input_index)
+ parse_conv(
+ self.inputs[input_index].conv,
+ input_layer.name,
+ self.config.inputs[input_index].conv_conf,
+ num_filters,
+ trans=True)
conv_conf = self.config.inputs[input_index].conv_conf
- parse_conv3d(
- self.inputs[input_index].conv, input_layer.name, conv_conf,
- num_filters
- ) # for z-axis pad:0, strid:1, filter_size:1, img_size:1
psize = self.calc_parameter_size(conv_conf)
self.create_input_parameter(input_index, psize)
- self.set_cnn_layer(name, conv_conf.output_z, conv_conf.output_y,
- conv_conf.output_x, self.config.num_filters)
+ self.set_cnn_layer(name, conv_conf.img_size_y, conv_conf.img_size,
+ self.config.num_filters)
psize = self.config.size
if shared_biases:
@@ -2048,62 +2051,42 @@ class Conv3DLayerBase(LayerBase):
self.create_bias_parameter(bias, psize, [psize, 1])
def calc_parameter_size(self, conv_conf):
- return self.config.num_filters * conv_conf.filter_channels \
- * (conv_conf.filter_size * conv_conf.filter_size_y \
- * conv_conf.filter_size_z)
+ return conv_conf.channels * conv_conf.filter_channels \
+ * (conv_conf.filter_size * conv_conf.filter_size_y)
- def set_layer_height_width(self, depth, height, width):
- self.config.depth = depth
- self.config.height = height
- self.config.width = width
- def set_cnn_layer(self,
- input_layer_name,
- depth,
- height,
- width,
- channels,
- is_print=True):
- size = depth * height * width * channels
- self.set_layer_size(size)
- self.set_layer_height_width(depth, height, width)
- if is_print:
- print("output for %s: c = %d, d = %d, h = %d, w = %d, size = %d" %
- (input_layer_name, channels, depth, height, width, size))
+@config_layer('exconvt')
+class ConvTransLayer(ConvTransLayerBase):
+ layer_type = 'exconvt'
-@config_layer('conv3d')
-class Conv3DLayer(Conv3DLayerBase):
- layer_type = 'conv3d'
+@config_layer('cudnn_convt')
+class ConvTransLayer(ConvTransLayerBase):
+ layer_type = 'cudnn_convt'
-@config_layer('convt_3d')
-class Conv3DTransLayerBase(LayerBase):
+@config_layer('conv_3d')
+class Conv3DLayerBase(LayerBase):
def __init__(self,
name,
inputs=[],
bias=True,
num_filters=None,
- shared_biases=False,
+ shared_biases=True,
**xargs):
- super(Conv3DTransLayerBase, self).__init__(
+ super(Conv3DLayerBase, self).__init__(
name, self.layer_type, 0, inputs=inputs, **xargs)
if num_filters is not None:
self.config.num_filters = num_filters
- use_gpu = int(g_command_config_args.get("use_gpu", 0))
- parallel_nn = int(g_command_config_args.get("parallel_nn", 0))
-
- # Automatically select cudnn_type for GPU and exconv for CPU
- # if set type=conv, but still reserve the way user specify
- # exconv or cudnn_conv manually.
- if self.layer_type == "cudnn_deconv3d":
- config_assert(use_gpu, "cudnn_conv3d only support GPU")
-
# need to specify layer in config
self.config.type = self.layer_type
+ trans = False
+ if self.config.type == "deconv3d":
+ trans = True
+
if shared_biases is not None:
self.config.shared_biases = shared_biases
@@ -2115,12 +2098,17 @@ class Conv3DTransLayerBase(LayerBase):
input_layer.name,
conv_conf,
num_filters,
- trans=True
+ trans=trans
) # for z-axis pad:0, strid:1, filter_size:1, img_size:1
psize = self.calc_parameter_size(conv_conf)
self.create_input_parameter(input_index, psize)
- self.set_cnn_layer(name, conv_conf.img_size_z, conv_conf.img_size_y,
- conv_conf.img_size, self.config.num_filters)
+ if trans:
+ self.set_cnn_layer(name, conv_conf.img_size_z,
+ conv_conf.img_size_y, conv_conf.img_size,
+ self.config.num_filters)
+ else:
+ self.set_cnn_layer(name, conv_conf.output_z, conv_conf.output_y,
+ conv_conf.output_x, self.config.num_filters)
psize = self.config.size
if shared_biases:
@@ -2132,11 +2120,6 @@ class Conv3DTransLayerBase(LayerBase):
* (conv_conf.filter_size * conv_conf.filter_size_y \
* conv_conf.filter_size_z)
- def set_layer_height_width(self, depth, height, width):
- self.config.depth = depth
- self.config.height = height
- self.config.width = width
-
def set_cnn_layer(self,
input_layer_name,
depth,
@@ -2146,86 +2129,21 @@ class Conv3DTransLayerBase(LayerBase):
is_print=True):
size = depth * height * width * channels
self.set_layer_size(size)
- self.set_layer_height_width(depth, height, width)
+ self.set_layer_height_width(height, width)
+ self.set_layer_depth(depth)
if is_print:
print("output for %s: c = %d, d = %d, h = %d, w = %d, size = %d" %
(input_layer_name, channels, depth, height, width, size))
-@config_layer('deconv3d')
-class DeConv3DLayer(Conv3DTransLayerBase):
- layer_type = 'deconv3d'
-
-
-@config_layer('convt')
-class ConvTransLayerBase(LayerBase):
- layer_type = 'convt'
-
- def __init__(self,
- name,
- inputs=[],
- bias=True,
- num_filters=None,
- shared_biases=False,
- **xargs):
- super(ConvTransLayerBase, self).__init__(
- name, self.layer_type, 0, inputs=inputs, **xargs)
-
- if num_filters is not None:
- self.config.num_filters = num_filters
-
- use_gpu = int(g_command_config_args.get("use_gpu", 0))
- parallel_nn = int(g_command_config_args.get("parallel_nn", 0))
-
- # Automatically select cudnn_type for GPU and exconvt for CPU
- # if set type=exconvt, but still reserve the way user specify
- # exconvt or cudnn_convt manually.
- if self.layer_type == "cudnn_convt":
- config_assert(use_gpu, "cudnn_convt only support GPU")
-
- if (use_gpu == 1 and self.layer_type != "exconvt" and
- (parallel_nn == 0 or self.config.device > -1)):
- self.layer_type = "cudnn_convt"
- else:
- self.layer_type = "exconvt"
- # need to specify layer in config
- self.config.type = self.layer_type
-
- if shared_biases is not None:
- self.config.shared_biases = shared_biases
-
- for input_index in xrange(len(self.inputs)):
- input_layer = self.get_input_layer(input_index)
- parse_conv(
- self.inputs[input_index].conv,
- input_layer.name,
- self.config.inputs[input_index].conv_conf,
- num_filters,
- trans=True)
- conv_conf = self.config.inputs[input_index].conv_conf
- psize = self.calc_parameter_size(conv_conf)
- self.create_input_parameter(input_index, psize)
- self.set_cnn_layer(name, conv_conf.img_size_y, conv_conf.img_size,
- self.config.num_filters)
-
- psize = self.config.size
- if shared_biases:
- psize = self.config.num_filters
- self.create_bias_parameter(bias, psize, [psize, 1])
-
- def calc_parameter_size(self, conv_conf):
- return conv_conf.channels * conv_conf.filter_channels \
- * (conv_conf.filter_size * conv_conf.filter_size_y)
-
-
-@config_layer('exconvt')
-class ConvTransLayer(ConvTransLayerBase):
- layer_type = 'exconvt'
+@config_layer('conv3d')
+class Conv3DLayer(Conv3DLayerBase):
+ layer_type = 'conv3d'
-@config_layer('cudnn_convt')
-class ConvTransLayer(ConvTransLayerBase):
- layer_type = 'cudnn_convt'
+@config_layer('deconv3d')
+class Conv3DLayer(Conv3DLayerBase):
+ layer_type = 'deconv3d'
@config_layer('norm')
diff --git a/python/paddle/trainer_config_helpers/layers.py b/python/paddle/trainer_config_helpers/layers.py
index 6953f134c5..e3ae81459f 100644
--- a/python/paddle/trainer_config_helpers/layers.py
+++ b/python/paddle/trainer_config_helpers/layers.py
@@ -6161,12 +6161,6 @@ def img_conv3d_layer(input,
param_attr=None,
shared_biases=True,
layer_attr=None,
- filter_size_y=None,
- stride_y=None,
- padding_y=None,
- filter_size_z=None,
- stride_z=None,
- padding_z=None,
trans=False,
layer_type=None):
"""
@@ -6175,7 +6169,7 @@ def img_conv3d_layer(input,
.. code-block:: python
- conv = img_conv3d_layer(input=data, filter_size=1, filter_size_y=1,
+ conv = img_conv3d_layer(input=data, filter_size=1,
num_channels=8,
num_filters=16, stride=1,
bias_attr=False,
@@ -6185,13 +6179,8 @@ def img_conv3d_layer(input,
:type name: basestring
:param input: Layer Input.
:type input: LayerOutput
- :param filter_size: The x dimension of a filter kernel. Or input a tuple for
- two image dimension.
+ :param filter_size: The x dimension of a filter kernel. Or input a list.
:type filter_size: int|tuple|list
- :param filter_size_y: The y dimension of a filter kernel. Since PaddlePaddle
- currently supports rectangular filters, the filter's
- shape will be (filter_size, filter_size_y).
- :type filter_size_y: int|None
:param num_filters: Each filter group's number of filter
:param act: Activation type. Default is tanh
:type act: BaseActivation
@@ -6200,13 +6189,9 @@ def img_conv3d_layer(input,
:param stride: The x dimension of the stride. Or input a tuple for two image
dimension.
:type stride: int|tuple|list
- :param stride_y: The y dimension of the stride.
- :type stride_y: int
:param padding: The x dimension of the padding. Or input a tuple for two
image dimension
:type padding: int|tuple|list
- :param padding_y: The y dimension of the padding.
- :type padding_y: int
:param bias_attr: Convolution bias attribute. None means default bias.
False means no bias.
:type bias_attr: ParameterAttribute|False
@@ -6233,47 +6218,26 @@ def img_conv3d_layer(input,
assert input.num_filters is not None
num_channels = input.num_filters
- if filter_size_y is None:
- if isinstance(filter_size, collections.Sequence):
- assert len(filter_size) == 2
- filter_size, filter_size_y = filter_size
- else:
- filter_size_y = filter_size
-
- if filter_size_z is None:
- if isinstance(filter_size, collections.Sequence):
- assert len(filter_size) == 2
- filter_size, filter_size_z = filter_size
- else:
- filter_size_z = filter_size
-
- if stride_y is None:
- if isinstance(stride, collections.Sequence):
- assert len(stride) == 2
- stride, stride_y = stride
- else:
- stride_y = stride
-
- if stride_z is None:
- if isinstance(stride, collections.Sequence):
- assert len(stride) == 2
- stride, stride_z = stride
- else:
- stride_z = stride
+ if isinstance(filter_size, collections.Sequence):
+ assert len(filter_size) == 3
+ filter_size, filter_size_y, filter_size_z = filter_size
+ else:
+ filter_size_y = filter_size
+ filter_size_z = filter_size
- if padding_y is None:
- if isinstance(padding, collections.Sequence):
- assert len(padding) == 2
- padding, padding_y = padding
- else:
- padding_y = padding
+ if isinstance(stride, collections.Sequence):
+ assert len(stride) == 3
+ stride, stride_y, stride_z = stride
+ else:
+ stride_y = stride
+ stride_z = stride
- if padding_z is None:
- if isinstance(padding, collections.Sequence):
- assert len(padding) == 2
- padding, padding_z = padding
- else:
- padding_z = padding
+ if isinstance(padding, collections.Sequence):
+ assert len(padding) == 3
+ padding, padding_y, padding_z = padding
+ else:
+ padding_y = padding
+ padding_z = padding
if param_attr.attr.get('initial_smart'):
# special initial for conv layers.
diff --git a/python/paddle/trainer_config_helpers/tests/configs/conv3d_deconv3d_test_config.py b/python/paddle/trainer_config_helpers/tests/configs/conv3d_deconv3d_test_config.py
index da0d23d057..15f7c1d271 100644
--- a/python/paddle/trainer_config_helpers/tests/configs/conv3d_deconv3d_test_config.py
+++ b/python/paddle/trainer_config_helpers/tests/configs/conv3d_deconv3d_test_config.py
@@ -14,23 +14,44 @@ padding_y = 1
padding_z = 1
groups = 1
-data = data_layer(
- name='data1', size=12096 * num_channels, height=48, width=42, depth=6)
+data1 = data_layer(name='data1', size=2016 * num_channels, height=48, width=42)
-conv3d = img_conv3d_layer(
+img_conv_layer(
+ input=data1,
+ filter_size=filter_size,
+ num_channels=num_channels,
+ num_filters=16,
+ stride=stride,
+ padding=padding,
+ act=LinearActivation(),
+ bias_attr=False)
+
+data = data_layer(
+ name='data', size=12096 * num_channels, height=48, width=42, depth=6)
+# first
+conv3d_1 = img_conv3d_layer(
input=data,
name='conv3d_1',
num_filters=16,
num_channels=num_channels,
filter_size=filter_size,
- filter_size_y=filter_size,
- filter_size_z=filter_size,
stride=stride,
- stride_y=stride_y,
- stride_z=stride_z,
padding=padding,
- padding_y=padding_y,
- padding_z=padding_z,
+ groups=groups,
+ bias_attr=True,
+ shared_biases=True,
+ trans=False,
+ layer_type="conv3d",
+ act=LinearActivation())
+# second
+conv3d_2 = img_conv3d_layer(
+ input=data,
+ name='conv3d_2',
+ num_filters=16,
+ num_channels=num_channels,
+ filter_size=[filter_size, filter_size_y, filter_size_z],
+ stride=[stride, stride_y, stride_z],
+ padding=[padding, padding_y, padding_z],
groups=groups,
bias_attr=True,
shared_biases=True,
@@ -38,61 +59,33 @@ conv3d = img_conv3d_layer(
layer_type="conv3d",
act=LinearActivation())
-deconv3d = img_conv3d_layer(
+# first
+deconv3d_1 = img_conv3d_layer(
input=data,
name='deconv3d_1',
num_filters=16,
num_channels=num_channels,
filter_size=filter_size,
- filter_size_y=filter_size,
- filter_size_z=filter_size,
stride=stride,
- stride_y=stride_y,
- stride_z=stride_z,
padding=padding,
- padding_y=padding_y,
- padding_z=padding_z,
groups=groups,
bias_attr=True,
shared_biases=True,
- trans=True,
+ trans=False,
layer_type="deconv3d",
act=LinearActivation())
-
-data = data_layer(name="input", size=8 * 16 * 16)
-conv1 = img_conv_layer(
- input=data,
- filter_size=1,
- filter_size_y=1,
- num_channels=8,
- num_filters=16,
- stride=1,
- bias_attr=False,
- act=ReluActivation(),
- layer_type="exconv")
-conv2 = img_conv_layer(
- input=data,
- filter_size=1,
- filter_size_y=1,
- num_channels=8,
- num_filters=16,
- stride=1,
- bias_attr=False,
- act=ReluActivation(),
- layer_type="exconv")
-
-concat = concat_layer(input=[conv1, conv2])
-
-conv = img_conv_layer(
+# second
+deconv3d_2 = img_conv3d_layer(
input=data,
- filter_size=1,
- filter_size_y=1,
- num_channels=8,
+ name='deconv3d_2',
num_filters=16,
- stride=1,
+ num_channels=num_channels,
+ filter_size=[filter_size, filter_size_y, filter_size_z],
+ stride=[stride, stride_y, stride_z],
+ padding=[padding, padding_y, padding_z],
+ groups=groups,
bias_attr=True,
- act=LinearActivation(),
- groups=2,
- layer_type="exconv")
-
-outputs(concat, conv)
+ shared_biases=True,
+ trans=False,
+ layer_type="deconv3d",
+ act=LinearActivation())