m53297601
/
Paddle
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

move tree_conv to fluid.contrib.layers (#19918)

Browse Source 
* move tree_conv to fluid.contrib.layers

test=develop

* update API.spec for tree_conv

test=develop

* update tree_conv api to increase unit coverage

test=develop
expand_as_op_1
Tao Luo 6 years ago committed by GitHub
parent 30adea0a23
commit a4919d3688
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 68 additions and 72 deletions
Show Stats Download Patch File Download Diff File

2
paddle/fluid/API.spec
Unescape View File

@ -285,7 +285,6 @@ paddle.fluid.layers.prroi_pool (ArgSpec(args=['input', 'rois', 'output_channels'
paddle.fluid.layers.teacher_student_sigmoid_loss (ArgSpec(args=['input', 'label', 'soft_max_up_bound', 'soft_max_lower_bound'], varargs=None, keywords=None, defaults=(15.0, -15.0)), ('document', '07cb0d95a646dba1b9cc7cdce89e59f0'))
paddle.fluid.layers.huber_loss (ArgSpec(args=['input', 'label', 'delta'], varargs=None, keywords=None, defaults=None), ('document', '11bb8e62cc9256958eff3991fe4834da'))
paddle.fluid.layers.kldiv_loss (ArgSpec(args=['x', 'target', 'reduction', 'name'], varargs=None, keywords=None, defaults=('mean', None)), ('document', '18bc95c62d3300456c3c7da5278b47bb'))
paddle.fluid.layers.tree_conv (ArgSpec(args=['nodes_vector', 'edge_set', 'output_size', 'num_filters', 'max_depth', 'act', 'param_attr', 'bias_attr', 'name'], varargs=None, keywords=None, defaults=(1, 2, 'tanh', None, None, None)), ('document', '864f3cdc5e0c6152e2a39b136171644f'))
paddle.fluid.layers.npair_loss (ArgSpec(args=['anchor', 'positive', 'labels', 'l2_reg'], varargs=None, keywords=None, defaults=(0.002,)), ('document', '6b6ee1170fe20a79cf0631a1f49b0df2'))
paddle.fluid.layers.pixel_shuffle (ArgSpec(args=['x', 'upscale_factor'], varargs=None, keywords=None, defaults=None), ('document', '7e5cac851fd9bad344230e1044b6a565'))
paddle.fluid.layers.fsp_matrix (ArgSpec(args=['x', 'y'], varargs=None, keywords=None, defaults=None), ('document', '20992b20d19c2e5983f366150827b4a6'))
@ -519,6 +518,7 @@ paddle.fluid.contrib.fused_elemwise_activation (ArgSpec(args=['x', 'y', 'functor
paddle.fluid.contrib.sequence_topk_avg_pooling (ArgSpec(args=['input', 'row', 'col', 'topks', 'channel_num'], varargs=None, keywords=None, defaults=None), ('document', '5218c85dd4122b626da9bb92f3b50042'))
paddle.fluid.contrib.var_conv_2d (ArgSpec(args=['input', 'row', 'col', 'input_channel', 'output_channel', 'filter_size', 'stride', 'param_attr', 'act', 'dtype', 'name'], varargs=None, keywords=None, defaults=(1, None, None, 'float32', None)), ('document', 'f52a6edf6d3e970568788604da3329c2'))
paddle.fluid.contrib.match_matrix_tensor (ArgSpec(args=['x', 'y', 'channel_num', 'act', 'param_attr', 'dtype', 'name'], varargs=None, keywords=None, defaults=(None, None, 'float32', None)), ('document', '3bdc4b2891c1460bc630fdcd22766b21'))
paddle.fluid.contrib.tree_conv (ArgSpec(args=['nodes_vector', 'edge_set', 'output_size', 'num_filters', 'max_depth', 'act', 'param_attr', 'bias_attr', 'name'], varargs=None, keywords=None, defaults=(1, 2, 'tanh', None, None, None)), ('document', '7c727562ebdda38274106d1a9b338e5b'))
paddle.fluid.contrib.BasicGRUUnit ('paddle.fluid.contrib.layers.rnn_impl.BasicGRUUnit', ('document', '2aed2540ed1540f081be9f4d08f2a65e'))
paddle.fluid.contrib.BasicGRUUnit.__init__ (ArgSpec(args=['self', 'name_scope', 'hidden_size', 'param_attr', 'bias_attr', 'gate_activation', 'activation', 'dtype'], varargs=None, keywords=None, defaults=(None, None, None, None, 'float32')), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.contrib.BasicGRUUnit.add_parameter (ArgSpec(args=['self', 'name', 'parameter'], varargs=None, keywords=None, defaults=None), ('document', 'f35ab374c7d5165c3daf3bd64a5a2ec1'))

66
python/paddle/fluid/contrib/layers/nn.py
Unescape View File

@ -29,6 +29,7 @@ __all__ = [
'sequence_topk_avg_pooling',
'var_conv_2d',
'match_matrix_tensor',
'tree_conv',
]
@ -361,3 +362,68 @@ def sequence_topk_avg_pooling(input, row, col, topks, channel_num):
'channel_num': channel_num})
return out
def tree_conv(nodes_vector,
edge_set,
output_size,
num_filters=1,
max_depth=2,
act='tanh',
param_attr=None,
bias_attr=None,
name=None):
"""
${comment}
Args:
nodes_vector(${nodes_vector_type}): ${nodes_vector_comment}
edge_set(${edge_set_type}): ${edge_set_comment}
output_size(int): output feature width
num_filters(int): number of filters, Default 1
max_depth(int): max depth of filters, Default 2
act(str): activation function, Default tanh
param_attr(ParamAttr): the parameter attribute for the filters, Default None
bias_attr(ParamAttr): the parameter attribute for the bias of this layer, Default None
name(str): a name of this layer(optional). If set None, the layer will be named automatically, Default None
Returns:
out(${out_type}): ${out_comment}
Examples:
.. code-block:: python
import paddle.fluid as fluid
# 10 for max_node_size of dataset, 5 for vector width
nodes_vector = fluid.layers.data(name='vectors', shape=[10, 5], dtype='float32')
# 10 for max_node_size of dataset, 2 for every edge has two nodes
# edges must be directional
edge_set = fluid.layers.data(name='edge_set', shape=[10, 2], dtype='float32')
# the shape of output will be [10, 6, 1],
# 10 for max_node_size of dataset, 6 for output size, 1 for 1 filter
out_vector = fluid.layers.tree_conv(nodes_vector, edge_set, 6, 1, 2)
# After reshape, output tensor could be nodes_vector for next tree convolution
out_vector = fluid.layers.reshape(out_vector, shape=[-1, 10, 6])
out_vector_2 = fluid.layers.tree_conv(out_vector, edge_set, 3, 4, 2)
# also output tensor could be pooling(the pooling in paper called global pooling)
pooled = fluid.layers.reduce_max(out_vector, dim=2) # global pooling
"""
helper = LayerHelper("tree_conv", **locals())
dtype = helper.input_dtype('nodes_vector')
feature_size = nodes_vector.shape[2]
W_shape = [feature_size, 3, output_size, num_filters]
W = helper.create_parameter(
attr=param_attr, shape=W_shape, dtype=dtype, is_bias=False)
out = helper.create_variable_for_type_inference(dtype=dtype)
helper.append_op(
type='tree_conv',
inputs={'NodesVector': nodes_vector,
'EdgeSet': edge_set,
'Filter': W},
outputs={'Out': out, },
attrs={'max_depth': max_depth})
if helper.bias_attr:
pre_activation = helper.append_bias_op(out)
else:
pre_activation = out
return helper.append_activation(pre_activation)

70
python/paddle/fluid/layers/nn.py
Unescape View File

@ -208,7 +208,6 @@ __all__ = [
'teacher_student_sigmoid_loss',
'huber_loss',
'kldiv_loss',
'tree_conv',
'npair_loss',
'pixel_shuffle',
'fsp_matrix',
@ -13272,75 +13271,6 @@ def kldiv_loss(x, target, reduction='mean', name=None):
return loss
@templatedoc()
def tree_conv(nodes_vector,
edge_set,
output_size,
num_filters=1,
max_depth=2,
act='tanh',
param_attr=None,
bias_attr=None,
name=None):
"""
${comment}
Args:
nodes_vector(${nodes_vector_type}): ${nodes_vector_comment}
edge_set(${edge_set_type}): ${edge_set_comment}
output_size(int): output feature width
num_filters(int): number of filters, Default 1
max_depth(int): max depth of filters, Default 2
act(str): activation function, Default tanh
param_attr(ParamAttr): the parameter attribute for the filters, Default None
bias_attr(ParamAttr): the parameter attribute for the bias of this layer, Default None
name(str): a name of this layer(optional). If set None, the layer will be named automatically, Default None
Returns:
out(${out_type}): ${out_comment}
Examples:
.. code-block:: python
import paddle.fluid as fluid
# 10 for max_node_size of dataset, 5 for vector width
nodes_vector = fluid.layers.data(name='vectors', shape=[10, 5], dtype='float32')
# 10 for max_node_size of dataset, 2 for every edge has two nodes
# edges must be directional
edge_set = fluid.layers.data(name='edge_set', shape=[10, 2], dtype='float32')
# the shape of output will be [10, 6, 1],
# 10 for max_node_size of dataset, 6 for output size, 1 for 1 filter
out_vector = fluid.layers.tree_conv(nodes_vector, edge_set, 6, 1, 2)
# After reshape, output tensor could be nodes_vector for next tree convolution
out_vector = fluid.layers.reshape(out_vector, shape=[-1, 10, 6])
out_vector_2 = fluid.layers.tree_conv(out_vector, edge_set, 3, 4, 2)
# also output tensor could be pooling(the pooling in paper called global pooling)
pooled = fluid.layers.reduce_max(out_vector, dim=2) # global pooling
"""
helper = LayerHelper("tree_conv", **locals())
dtype = helper.input_dtype('nodes_vector')
feature_size = nodes_vector.shape[2]
W_shape = [feature_size, 3, output_size, num_filters]
W = helper.create_parameter(
attr=param_attr, shape=W_shape, dtype=dtype, is_bias=False)
if name == None:
out = helper.create_variable_for_type_inference(dtype=dtype)
else:
out = helper.create_variable(name=name, dtype=dtype, persistable=False)
helper.append_op(
type='tree_conv',
inputs={'NodesVector': nodes_vector,
'EdgeSet': edge_set,
'Filter': W},
outputs={'Out': out, },
attrs={'max_depth': max_depth})
if helper.bias_attr:
pre_activation = helper.append_bias_op(out)
else:
pre_activation = out
return helper.append_activation(pre_activation)
from .ops import square
from .control_flow import equal

2
python/paddle/fluid/tests/unittests/test_layers.py
Unescape View File

@ -876,7 +876,7 @@ class TestLayer(LayerTest):
dtype='int32',
lod_level=1,
append_batch_size=False)
ret = layers.tree_conv(
ret = fluid.contrib.layers.tree_conv(
nodes_vector=NodesVector,
edge_set=EdgeSet,
output_size=6,

Write Preview
Loading…
Cancel
Save