Skip to content

B
book

PaddlePaddle / book

通知 16
Star 4
Fork 0
B
book
提交 9e6a24cd 编写于 作者: G gongweibao
浏览文件
操作

Merge remote-tracking branch 'upstream/develop' into develop

上级 c2c24bc8 f3b57895
展开全部 隐藏空白更改
内联 并排
Showing with 2358 addition and 1785 deletion
fit_a_line/README.md
浏览文件 @ 9e6a24cd
......@@ -39,20 +39,32 @@ $$MSE=\frac{1}{n}\sum_{i=1}^{n}{(\hat{Y_i}-Y_i)}^2$$
### 训练过程
定义好模型结构之后，我们要通过以下几个步骤进行模型训练
1. 初始化参数，其中包括权重$\omega_i$和偏置$b$，对其进行初始化（如0均值，1方差）。
2. 网络正向传播计算网络输出和损失函数。
3. 根据损失函数进行反向误差传播 （[backpropagation](https://en.wikipedia.org/wiki/Backpropagation)），将网络误差从输出层依次向前传递, 并更新网络中的参数。
4. 重复2~3步骤，直至网络训练误差达到规定的程度或训练轮次达到设定值。
定义好模型结构之后，我们要通过以下几个步骤进行模型训练
1. 初始化参数，其中包括权重$\omega_i$和偏置$b$，对其进行初始化（如0均值，1方差）。
2. 网络正向传播计算网络输出和损失函数。
3. 根据损失函数进行反向误差传播 （[backpropagation](https://en.wikipedia.org/wiki/Backpropagation)），将网络误差从输出层依次向前传递, 并更新网络中的参数。
4. 重复2~3步骤，直至网络训练误差达到规定的程度或训练轮次达到设定值。
## 数据集
### 数据集接口的封装
首先加载需要的包
## 数据准备
执行以下命令来准备数据:
```bash
cd data && python prepare_data.py
```python
import paddle.v2 as paddle
import paddle.v2.dataset.uci_housing as uci_housing
```
这段代码将从[UCI Housing Data Set](https://archive.ics.uci.edu/ml/datasets/Housing)下载数据并进行[预处理](#数据预处理)，最后数据将被分为训练集和测试集。
我们通过uci_housing模块引入了数据集合[UCI Housing Data Set](https://archive.ics.uci.edu/ml/datasets/Housing)
其中，在uci_housing模块中封装了：
1. 数据下载的过程<br>
下载数据保存在~/.cache/paddle/dataset/uci_housing/housing.data<br>
2. [数据预处理](#数据预处理)的过程<br>
### 数据集介绍
这份数据集共506行，每行包含了波士顿郊区的一类房屋的相关信息及该类房屋价格的中位数。其各维属性的意义如下：
| 属性名 | 解释 | 类型 |
......@@ -90,89 +102,94 @@ cd data && python prepare_data.py
</p>
#### 整理训练集与测试集
我们将数据集分割为两份：一份用于调整模型的参数，即进行模型的训练，模型在这份数据集上的误差被称为**训练误差**；另外一份被用来测试，模型在这份数据集上的误差被称为**测试误差**。我们训练模型的目的是为了通过从训练数据中找到规律来预测未知的新数据，所以测试误差是更能反映模型表现的指标。分割数据的比例要考虑到两个因素：更多的训练数据会降低参数估计的方差，从而得到更可信的模型；而更多的测试数据会降低测试误差的方差，从而得到更可信的测试误差。一种常见的分割比例为$8:2$，感兴趣的读者朋友们也可以尝试不同的设置来观察这两种误差的变化。
我们将数据集分割为两份：一份用于调整模型的参数，即进行模型的训练，模型在这份数据集上的误差被称为**训练误差**；另外一份被用来测试，模型在这份数据集上的误差被称为**测试误差**。我们训练模型的目的是为了通过从训练数据中找到规律来预测未知的新数据，所以测试误差是更能反映模型表现的指标。分割数据的比例要考虑到两个因素：更多的训练数据会降低参数估计的方差，从而得到更可信的模型；而更多的测试数据会降低测试误差的方差，从而得到更可信的测试误差。我们这个例子中设置的分割比例为$8:2$
执行如下命令可以分割数据集，并将训练集和测试集的地址分别写入train.list 和 test.list两个文件中，供PaddlePaddle读取。
```python
python prepare_data.py -r 0.8 #默认使用8:2的比例进行分割
```
在更复杂的模型训练过程中，我们往往还会多使用一种数据集：验证集。因为复杂的模型中常常还有一些超参数（[Hyperparameter](https://en.wikipedia.org/wiki/Hyperparameter_optimization)）需要调节，所以我们会尝试多种超参数的组合来分别训练多个模型，然后对比它们在验证集上的表现选择相对最好的一组超参数，最后才使用这组参数下训练的模型在测试集上评估测试误差。由于本章训练的模型比较简单，我们暂且忽略掉这个过程。
### 提供数据给PaddlePaddle
准备好数据之后，我们使用一个Python data provider来为PaddlePaddle的训练过程提供数据。一个 data provider 就是一个Python函数，它会被PaddlePaddle的训练过程调用。在这个例子里，只需要读取已经保存好的数据，然后一行一行地返回给PaddlePaddle的训练进程即可。
## 训练
fit_a_line下trainer.py演示了训练的整体过程
```python
from paddle.trainer.PyDataProvider2 import *
import numpy as np
#定义数据的类型和维度
@provider(input_types=[dense_vector(13), dense_vector(1)])
def process(settings, input_file):
data = np.load(input_file.strip())
for row in data:
yield row[:-1].tolist(), row[-1:].tolist()
### 初始化paddlepaddle
```python
# init
paddle.init(use_gpu=False, trainer_count=1)
```
## 模型配置说明
### 模型配置
使用`fc_layer`和`LinearActivation`来表示线性回归的模型本身。
### 数据定义
首先，通过 `define_py_data_sources2` 来配置PaddlePaddle从上面的`dataprovider.py`里读入训练数据和测试数据。 PaddlePaddle接受从命令行读入的配置信息，例如这里我们传入一个名为`is_predict`的变量来控制模型在训练和测试时的不同结构。
```python
from paddle.trainer_config_helpers import *
#输入数据，13维的房屋信息
x = paddle.layer.data(name='x', type=paddle.data_type.dense_vector(13))
y_predict = paddle.layer.fc(input=x,
size=1,
act=paddle.activation.Linear())
y = paddle.layer.data(name='y', type=paddle.data_type.dense_vector(1))
cost = paddle.layer.regression_cost(input=y_predict, label=y)
```
### 创建参数
```python
# create parameters
parameters = paddle.parameters.create(cost)
```
is_predict = get_config_arg('is_predict', bool, False)
### 创建trainer
define_py_data_sources2(
train_list='data/train.list',
test_list='data/test.list',
module='dataprovider',
obj='process')
```python
# create optimizer
optimizer = paddle.optimizer.Momentum(momentum=0)
trainer = paddle.trainer.SGD(cost=cost,
parameters=parameters,
update_equation=optimizer)
```
### 算法配置
接着，指定模型优化算法的细节。由于线性回归模型比较简单，我们只要设置基本的`batch_size`即可，它指定每次更新参数的时候使用多少条数据计算梯度信息。
### 读取数据且打印训练的中间信息
在程序中，我们通过reader接口来获取训练或者测试的数据,通过eventhandler来打印训练的中间信息
feeding中设置了训练数据和测试数据的下标,reader通过下标区分训练和测试数据。
```python
settings(batch_size=2)
feeding={'x': 0,
'y': 1}
# event_handler to print training and testing info
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 100 == 0:
print "Pass %d, Batch %d, Cost %f" % (
event.pass_id, event.batch_id, event.cost)
if isinstance(event, paddle.event.EndPass):
result = trainer.test(
reader=paddle.batch(
uci_housing.test(), batch_size=2),
feeding=feeding)
print "Test %d, Cost %f" % (event.pass_id, result.cost)
```
### 开始训练
### 网络结构
最后，使用`fc_layer`和`LinearActivation`来表示线性回归的模型本身。
```python
#输入数据，13维的房屋信息
x = data_layer(name='x', size=13)
y_predict = fc_layer(
input=x,
param_attr=ParamAttr(name='w'),
size=1,
act=LinearActivation(),
bias_attr=ParamAttr(name='b'))
if not is_predict: #训练时，我们使用MSE，即regression_cost作为损失函数
y = data_layer(name='y', size=1)
cost = regression_cost(input=y_predict, label=y)
outputs(cost) #训练时输出MSE来监控损失的变化
else: #测试时，输出预测值
outputs(y_predict)
# training
trainer.train(
reader=paddle.batch(
paddle.reader.shuffle(
uci_housing.train(), buf_size=500),
batch_size=2),
feeding=feeding,
event_handler=event_handler,
num_passes=30)
```
## 训练模型
在对应代码的根目录下执行PaddlePaddle的命令行训练程序。这里指定模型配置文件为`trainer_config.py`，训练30轮，结果保存在`output`路径下。
```bash
./train.sh
```
## bash中执行训练程序
**注意设置好paddle的安装包路径**
## 应用模型
现在来看下如何使用已经训练好的模型进行预测。
```bash
python predict.py
```
这里默认使用`output/pass-00029`中保存的模型进行预测，并将数据中的房价与预测结果进行对比，结果保存在 `predictions.png`中。
如果你想使用别的模型或者其它的数据进行预测，只要传入新的路径即可：
```bash
python predict.py -m output/pass-00020 -t data/housing.test.npy
python train.py
```
## 总结
......
fit_a_line/train.py 0 → 100644
浏览文件 @ 9e6a24cd
import paddle.v2 as paddle
import paddle.v2.dataset.uci_housing as uci_housing
def main():
# init
paddle.init(use_gpu=False, trainer_count=1)
# network config
x = paddle.layer.data(name='x', type=paddle.data_type.dense_vector(13))
y_predict = paddle.layer.fc(input=x, size=1, act=paddle.activation.Linear())
y = paddle.layer.data(name='y', type=paddle.data_type.dense_vector(1))
cost = paddle.layer.regression_cost(input=y_predict, label=y)
# create parameters
parameters = paddle.parameters.create(cost)
# create optimizer
optimizer = paddle.optimizer.Momentum(momentum=0)
trainer = paddle.trainer.SGD(cost=cost,
parameters=parameters,
update_equation=optimizer)
feeding = {'x': 0, 'y': 1}
# event_handler to print training and testing info
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 100 == 0:
print "Pass %d, Batch %d, Cost %f" % (
event.pass_id, event.batch_id, event.cost)
if isinstance(event, paddle.event.EndPass):
result = trainer.test(
reader=paddle.batch(
uci_housing.test(), batch_size=2),
feeding=feeding)
print "Test %d, Cost %f" % (event.pass_id, result.cost)
# training
trainer.train(
reader=paddle.batch(
paddle.reader.shuffle(
uci_housing.train(), buf_size=500),
batch_size=2),
feeding=feeding,
event_handler=event_handler,
num_passes=30)
if __name__ == '__main__':
main()
image_classification/README.md
浏览文件 @ 9e6a24cd
此差异已折叠。
image_classification/deprecated/README.md 0 → 100644
浏览文件 @ 9e6a24cd
此差异已折叠。
image_classification/index.html
浏览文件 @ 9e6a24cd
此差异已折叠。
image_classification/resnet.py 0 → 100644
浏览文件 @ 9e6a24cd
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2 as paddle
__all__ = ['resnet_cifar10']
def conv_bn_layer(input,
ch_out,
filter_size,
stride,
padding,
active_type=paddle.activation.Relu(),
ch_in=None):
tmp = paddle.layer.img_conv(
input=input,
filter_size=filter_size,
num_channels=ch_in,
num_filters=ch_out,
stride=stride,
padding=padding,
act=paddle.activation.Linear(),
bias_attr=False)
return paddle.layer.batch_norm(input=tmp, act=active_type)
def shortcut(ipt, n_in, n_out, stride):
if n_in != n_out:
return conv_bn_layer(ipt, n_out, 1, stride, 0,
paddle.activation.Linear())
else:
return ipt
def basicblock(ipt, ch_out, stride):
ch_in = ch_out * 2
tmp = conv_bn_layer(ipt, ch_out, 3, stride, 1)
tmp = conv_bn_layer(tmp, ch_out, 3, 1, 1, paddle.activation.Linear())
short = shortcut(ipt, ch_in, ch_out, stride)
return paddle.layer.addto(input=[tmp, short], act=paddle.activation.Relu())
def layer_warp(block_func, ipt, features, count, stride):
tmp = block_func(ipt, features, stride)
for i in range(1, count):
tmp = block_func(tmp, features, 1)
return tmp
def resnet_cifar10(ipt, depth=32):
# depth should be one of 20, 32, 44, 56, 110, 1202
assert (depth - 2) % 6 == 0
n = (depth - 2) / 6
nStages = {16, 64, 128}
conv1 = conv_bn_layer(
ipt, ch_in=3, ch_out=16, filter_size=3, stride=1, padding=1)
res1 = layer_warp(basicblock, conv1, 16, n, 1)
res2 = layer_warp(basicblock, res1, 32, n, 2)
res3 = layer_warp(basicblock, res2, 64, n, 2)
pool = paddle.layer.img_pool(
input=res3, pool_size=8, stride=1, pool_type=paddle.pooling.Avg())
return pool
image_classification/train.py 0 → 100644
浏览文件 @ 9e6a24cd
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License
import sys
import paddle.v2 as paddle
from vgg import vgg_bn_drop
from resnet import resnet_cifar10
def main():
datadim = 3 * 32 * 32
classdim = 10
# PaddlePaddle init
paddle.init(use_gpu=False, trainer_count=1)
image = paddle.layer.data(
name="image", type=paddle.data_type.dense_vector(datadim))
# Add neural network config
# option 1. resnet
# net = resnet_cifar10(image, depth=32)
# option 2. vgg
net = vgg_bn_drop(image)
out = paddle.layer.fc(input=net,
size=classdim,
act=paddle.activation.Softmax())
lbl = paddle.layer.data(
name="label", type=paddle.data_type.integer_value(classdim))
cost = paddle.layer.classification_cost(input=out, label=lbl)
# Create parameters
parameters = paddle.parameters.create(cost)
# Create optimizer
momentum_optimizer = paddle.optimizer.Momentum(
momentum=0.9,
regularization=paddle.optimizer.L2Regularization(rate=0.0002 * 128),
learning_rate=0.1 / 128.0,
learning_rate_decay_a=0.1,
learning_rate_decay_b=50000 * 100,
learning_rate_schedule='discexp',
batch_size=128)
# End batch and end pass event handler
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 100 == 0:
print "\nPass %d, Batch %d, Cost %f, %s" % (
event.pass_id, event.batch_id, event.cost, event.metrics)
else:
sys.stdout.write('.')
sys.stdout.flush()
if isinstance(event, paddle.event.EndPass):
result = trainer.test(
reader=paddle.batch(
paddle.dataset.cifar.test10(), batch_size=128),
feeding={'image': 0,
'label': 1})
print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
# Create trainer
trainer = paddle.trainer.SGD(cost=cost,
parameters=parameters,
update_equation=momentum_optimizer)
trainer.train(
reader=paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10(), buf_size=50000),
batch_size=128),
num_passes=200,
event_handler=event_handler,
feeding={'image': 0,
'label': 1})
if __name__ == '__main__':
main()
image_classification/vgg.py 0 → 100644
浏览文件 @ 9e6a24cd
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2 as paddle
__all__ = ['vgg_bn_drop']
def vgg_bn_drop(input):
def conv_block(ipt, num_filter, groups, dropouts, num_channels=None):
return paddle.networks.img_conv_group(
input=ipt,
num_channels=num_channels,
pool_size=2,
pool_stride=2,
conv_num_filter=[num_filter] * groups,
conv_filter_size=3,
conv_act=paddle.activation.Relu(),
conv_with_batchnorm=True,
conv_batchnorm_drop_rate=dropouts,
pool_type=paddle.pooling.Max())
conv1 = conv_block(input, 64, 2, [0.3, 0], 3)
conv2 = conv_block(conv1, 128, 2, [0.4, 0])
conv3 = conv_block(conv2, 256, 3, [0.4, 0.4, 0])
conv4 = conv_block(conv3, 512, 3, [0.4, 0.4, 0])
conv5 = conv_block(conv4, 512, 3, [0.4, 0.4, 0])
drop = paddle.layer.dropout(input=conv5, dropout_rate=0.5)
fc1 = paddle.layer.fc(input=drop, size=512, act=paddle.activation.Linear())
bn = paddle.layer.batch_norm(
input=fc1,
act=paddle.activation.Relu(),
layer_attr=paddle.attr.Extra(drop_rate=0.5))
fc2 = paddle.layer.fc(input=bn, size=512, act=paddle.activation.Linear())
return fc2
label_semantic_roles/README.en.md
浏览文件 @ 9e6a24cd
......@@ -225,6 +225,8 @@ We trained in the English Wikipedia language model to get a word vector lookup t
Get dictionary, print dictionary size:
```python
import math
import numpy as np
import paddle.v2 as paddle
import paddle.v2.dataset.conll05 as conll05
......@@ -233,164 +235,164 @@ word_dict_len = len(word_dict)
label_dict_len = len(label_dict)
pred_len = len(verb_dict)
print len(word_dict_len)
print len(label_dict_len)
print len(pred_len)
print word_dict_len
print label_dict_len
print pred_len
```
## Model configuration
1. Define input data dimensions and model hyperparameters.
- 1. Define input data dimensions and model hyperparameters.
```python
mark_dict_len = 2 # Value range of region mark. Region mark is either 0 or 1, so range is 2
word_dim = 32 # word vector dimension
mark_dim = 5 # adjacent dimension
hidden_dim = 512 # the dimension of LSTM hidden layer vector is 128 (512/4)
depth = 8 # depth of stacked LSTM
# There are 9 features per sample, so we will define 9 data layers.
# They type for each layer is integer_value_sequence.
def d_type(value_range):
return paddle.data_type.integer_value_sequence(value_range)
# word sequence
word = paddle.layer.data(name='word_data', type=d_type(word_dict_len))
# predicate
predicate = paddle.layer.data(name='verb_data', type=d_type(pred_len))
# 5 features for predicate context
ctx_n2 = paddle.layer.data(name='ctx_n2_data', type=d_type(word_dict_len))
ctx_n1 = paddle.layer.data(name='ctx_n1_data', type=d_type(word_dict_len))
ctx_0 = paddle.layer.data(name='ctx_0_data', type=d_type(word_dict_len))
ctx_p1 = paddle.layer.data(name='ctx_p1_data', type=d_type(word_dict_len))
ctx_p2 = paddle.layer.data(name='ctx_p2_data', type=d_type(word_dict_len))
# region marker sequence
mark = paddle.layer.data(name='mark_data', type=d_type(mark_dict_len))
# label sequence
target = paddle.layer.data(name='target', type=d_type(label_dict_len))
```
```python
mark_dict_len = 2 # Value range of region mark. Region mark is either 0 or 1, so range is 2
word_dim = 32 # word vector dimension
mark_dim = 5 # adjacent dimension
hidden_dim = 512 # the dimension of LSTM hidden layer vector is 128 (512/4)
depth = 8 # depth of stacked LSTM
# There are 9 features per sample, so we will define 9 data layers.
# They type for each layer is integer_value_sequence.
def d_type(value_range):
return paddle.data_type.integer_value_sequence(value_range)
# word sequence
word = paddle.layer.data(name='word_data', type=d_type(word_dict_len))
# predicate
predicate = paddle.layer.data(name='verb_data', type=d_type(pred_len))
# 5 features for predicate context
ctx_n2 = paddle.layer.data(name='ctx_n2_data', type=d_type(word_dict_len))
ctx_n1 = paddle.layer.data(name='ctx_n1_data', type=d_type(word_dict_len))
ctx_0 = paddle.layer.data(name='ctx_0_data', type=d_type(word_dict_len))
ctx_p1 = paddle.layer.data(name='ctx_p1_data', type=d_type(word_dict_len))
ctx_p2 = paddle.layer.data(name='ctx_p2_data', type=d_type(word_dict_len))
# region marker sequence
mark = paddle.layer.data(name='mark_data', type=d_type(mark_dict_len))
# label sequence
target = paddle.layer.data(name='target', type=d_type(label_dict_len))
```
Speciala note: hidden_dim = 512 means LSTM hidden vector of 128 dimension (512/4). Please refer PaddlePaddle official documentation for detail: [lstmemory](http://www.paddlepaddle.org/doc/ui/api/trainer_config_helpers/layers.html#lstmemory)。
Speciala note: hidden_dim = 512 means LSTM hidden vector of 128 dimension (512/4). Please refer PaddlePaddle official documentation for detail: [lstmemory](http://www.paddlepaddle.org/doc/ui/api/trainer_config_helpers/layers.html#lstmemory)。
- 2. The word sequence, predicate, predicate context, and region mark sequence are transformed into embedding vector sequences.
```python
# Since word vectorlookup table is pre-trained, we won't update it this time.
# is_static being True prevents updating the lookup table during training.
emb_para = paddle.attr.Param(name='emb', initial_std=0., is_static=True)
# hyperparameter configurations
default_std = 1 / math.sqrt(hidden_dim) / 3.0
std_default = paddle.attr.Param(initial_std=default_std)
std_0 = paddle.attr.Param(initial_std=0.)
predicate_embedding = paddle.layer.embedding(
size=word_dim,
input=predicate,
param_attr=paddle.attr.Param(
name='vemb', initial_std=default_std))
mark_embedding = paddle.layer.embedding(
size=mark_dim, input=mark, param_attr=std_0)
word_input = [word, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2]
emb_layers = [
paddle.layer.embedding(
size=word_dim, input=x, param_attr=emb_para) for x in word_input
]
emb_layers.append(predicate_embedding)
emb_layers.append(mark_embedding)
```
2. The word sequence, predicate, predicate context, and region mark sequence are transformed into embedding vector sequences.
- 3. 8 LSTM units will be trained in "forward / backward" order.
```python
# Since word vectorlookup table is pre-trained, we won't update it this time.
# is_static being True prevents updating the lookup table during training.
emb_para = paddle.attr.Param(name='emb', initial_std=0., is_static=True)
# hyperparameter configurations
default_std = 1 / math.sqrt(hidden_dim) / 3.0
std_default = paddle.attr.Param(initial_std=default_std)
std_0 = paddle.attr.Param(initial_std=0.)
predicate_embedding = paddle.layer.embedding(
size=word_dim,
input=predicate,
param_attr=paddle.attr.Param(
name='vemb', initial_std=default_std))
mark_embedding = paddle.layer.embedding(
size=mark_dim, input=mark, param_attr=std_0)
word_input = [word, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2]
emb_layers = [
paddle.layer.embedding(
size=word_dim, input=x, param_attr=emb_para) for x in word_input
]
emb_layers.append(predicate_embedding)
emb_layers.append(mark_embedding)
```
3. 8 LSTM units will be trained in "forward / backward" order.
```python
hidden_0 = paddle.layer.mixed(
```python
hidden_0 = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=emb, param_attr=std_default) for emb in emb_layers
])
mix_hidden_lr = 1e-3
lstm_para_attr = paddle.attr.Param(initial_std=0.0, learning_rate=1.0)
hidden_para_attr = paddle.attr.Param(
initial_std=default_std, learning_rate=mix_hidden_lr)
lstm_0 = paddle.layer.lstmemory(
input=hidden_0,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
bias_attr=std_0,
param_attr=lstm_para_attr)
# stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0]
for i in range(1, depth):
mix_hidden = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=emb, param_attr=std_default) for emb in emb_layers
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
])
mix_hidden_lr = 1e-3
lstm_para_attr = paddle.attr.Param(initial_std=0.0, learning_rate=1.0)
hidden_para_attr = paddle.attr.Param(
initial_std=default_std, learning_rate=mix_hidden_lr)
lstm_0 = paddle.layer.lstmemory(
input=hidden_0,
lstm = paddle.layer.lstmemory(
input=mix_hidden,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
reverse=((i % 2) == 1),
bias_attr=std_0,
param_attr=lstm_para_attr)
# stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0]
for i in range(1, depth):
mix_hidden = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
])
lstm = paddle.layer.lstmemory(
input=mix_hidden,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
reverse=((i % 2) == 1),
bias_attr=std_0,
param_attr=lstm_para_attr)
input_tmp = [mix_hidden, lstm]
```
4. We will concatenate the output of top LSTM unit with it's input, and project into a hidden layer. Then put a fully connected layer on top of it to get the final vector representation.
```python
feature_out = paddle.layer.mixed(
input_tmp = [mix_hidden, lstm]
```
- 4. We will concatenate the output of top LSTM unit with it's input, and project into a hidden layer. Then put a fully connected layer on top of it to get the final vector representation.
```python
feature_out = paddle.layer.mixed(
size=label_dict_len,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
], )
```
- 5. We use CRF as cost function, the parameter of CRF cost will be named `crfw`.
```python
crf_cost = paddle.layer.crf(
size=label_dict_len,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
], )
```
5. We use CRF as cost function, the parameter of CRF cost will be named `crfw`.
```python
crf_cost = paddle.layer.crf(
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(
name='crfw',
initial_std=default_std,
learning_rate=mix_hidden_lr))
```
6. CRF decoding layer is used for evaluation and inference. It shares parameter with CRF layer. The sharing of parameters among multiple layers is specified by the same parameter name in these layers.
```python
crf_dec = paddle.layer.crf_decoding(
name='crf_dec_l',
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(name='crfw'))
```
input=feature_out,
label=target,
param_attr=paddle.attr.Param(
name='crfw',
initial_std=default_std,
learning_rate=mix_hidden_lr))
```
- 6. CRF decoding layer is used for evaluation and inference. It shares parameter with CRF layer. The sharing of parameters among multiple layers is specified by the same parameter name in these layers.
```python
crf_dec = paddle.layer.crf_decoding(
name='crf_dec_l',
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(name='crfw'))
```
## Train model
......@@ -413,8 +415,8 @@ Now we load pre-trained word lookup table.
```python
def load_parameter(file_name, h, w):
with open(file_name, 'rb') as f:
f.read(16)
return np.fromfile(f, dtype=np.float32).reshape(h, w)
f.read(16)
return np.fromfile(f, dtype=np.float32).reshape(h, w)
parameters.set('emb', load_parameter(conll05.get_embedding(), 44068, 32))
```
......
label_semantic_roles/README.md
浏览文件 @ 9e6a24cd
......@@ -187,171 +187,175 @@ conll05st-release/
获取词典，打印词典大小：
```python
import math
import numpy as np
import paddle.v2 as paddle
import paddle.v2.dataset.conll05 as conll05
paddle.init(use_gpu=False, trainer_count=1)
word_dict, verb_dict, label_dict = conll05.get_dict()
word_dict_len = len(word_dict)
label_dict_len = len(label_dict)
pred_len = len(verb_dict)
print len(word_dict_len)
print len(label_dict_len)
print len(pred_len)
print word_dict_len
print label_dict_len
print pred_len
```
## 模型配置说明
1. 定义输入数据维度及模型超参数。
- 1. 定义输入数据维度及模型超参数。
```python
mark_dict_len = 2 # 谓上下文区域标志的维度，是一个0-1 2值特征，因此维度为2
word_dim = 32 # 词向量维度
mark_dim = 5 # 谓词上下文区域通过词表被映射为一个实向量，这个是相邻的维度
hidden_dim = 512 # LSTM隐层向量的维度 ： 512 / 4
depth = 8 # 栈式LSTM的深度
# 一条样本总共9个特征，下面定义了9个data层，每个层类型为integer_value_sequence，表示整数ID的序列类型.
def d_type(size):
return paddle.data_type.integer_value_sequence(size)
# 句子序列
word = paddle.layer.data(name='word_data', type=d_type(word_dict_len))
# 谓词
predicate = paddle.layer.data(name='verb_data', type=d_type(pred_len))
# 谓词上下文5个特征
ctx_n2 = paddle.layer.data(name='ctx_n2_data', type=d_type(word_dict_len))
ctx_n1 = paddle.layer.data(name='ctx_n1_data', type=d_type(word_dict_len))
ctx_0 = paddle.layer.data(name='ctx_0_data', type=d_type(word_dict_len))
ctx_p1 = paddle.layer.data(name='ctx_p1_data', type=d_type(word_dict_len))
ctx_p2 = paddle.layer.data(name='ctx_p2_data', type=d_type(word_dict_len))
# 谓词上下区域标志
mark = paddle.layer.data(name='mark_data', type=d_type(mark_dict_len))
```python
mark_dict_len = 2 # 谓上下文区域标志的维度，是一个0-1 2值特征，因此维度为2
word_dim = 32 # 词向量维度
mark_dim = 5 # 谓词上下文区域通过词表被映射为一个实向量，这个是相邻的维度
hidden_dim = 512 # LSTM隐层向量的维度 ： 512 / 4
depth = 8 # 栈式LSTM的深度
# 标注序列
target = paddle.layer.data(name='target', type=d_type(label_dict_len))
```
# 一条样本总共9个特征，下面定义了9个data层，每个层类型为integer_value_sequence，表示整数ID的序列类型.
def d_type(size):
return paddle.data_type.integer_value_sequence(size)
# 句子序列
word = paddle.layer.data(name='word_data', type=d_type(word_dict_len))
# 谓词
predicate = paddle.layer.data(name='verb_data', type=d_type(pred_len))
# 谓词上下文5个特征
ctx_n2 = paddle.layer.data(name='ctx_n2_data', type=d_type(word_dict_len))
ctx_n1 = paddle.layer.data(name='ctx_n1_data', type=d_type(word_dict_len))
ctx_0 = paddle.layer.data(name='ctx_0_data', type=d_type(word_dict_len))
ctx_p1 = paddle.layer.data(name='ctx_p1_data', type=d_type(word_dict_len))
ctx_p2 = paddle.layer.data(name='ctx_p2_data', type=d_type(word_dict_len))
# 谓词上下区域标志
mark = paddle.layer.data(name='mark_data', type=d_type(mark_dict_len))
# 标注序列
target = paddle.layer.data(name='target', type=d_type(label_dict_len))
```
这里需要特别说明的是hidden_dim = 512指定了LSTM隐层向量的维度为128维，关于这一点请参考PaddlePaddle官方文档中[lstmemory](http://www.paddlepaddle.org/doc/ui/api/trainer_config_helpers/layers.html#lstmemory)的说明。
2. 将句子序列、谓词、谓词上下文、谓词上下文区域标记通过词表，转换为实向量表示的词向量序列。
这里需要特别说明的是hidden_dim = 512指定了LSTM隐层向量的维度为128维，关于这一点请参考PaddlePaddle官方文档中[lstmemory](http://www.paddlepaddle.org/doc/ui/api/trainer_config_helpers/layers.html#lstmemory)的说明。
- 2. 将句子序列、谓词、谓词上下文、谓词上下文区域标记通过词表，转换为实向量表示的词向量序列。
```python
# 在本教程中，我们加载了预训练的词向量，这里设置了：is_static=True
# is_static 为 True 时保证了在训练 SRL 模型过程中，词表不再更新
emb_para = paddle.attr.Param(name='emb', initial_std=0., is_static=True)
# 设置超参数
default_std = 1 / math.sqrt(hidden_dim) / 3.0
std_default = paddle.attr.Param(initial_std=default_std)
std_0 = paddle.attr.Param(initial_std=0.)
predicate_embedding = paddle.layer.embedding(
size=word_dim,
input=predicate,
param_attr=paddle.attr.Param(
name='vemb', initial_std=default_std))
mark_embedding = paddle.layer.embedding(
size=mark_dim, input=mark, param_attr=std_0)
word_input = [word, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2]
emb_layers = [
paddle.layer.embedding(
size=word_dim, input=x, param_attr=emb_para) for x in word_input
]
emb_layers.append(predicate_embedding)
emb_layers.append(mark_embedding)
```
```python
# 在本教程中，我们加载了预训练的词向量，这里设置了：is_static=True
# is_static 为 True 时保证了在训练 SRL 模型过程中，词表不再更新
emb_para = paddle.attr.Param(name='emb', initial_std=0., is_static=True)
# 设置超参数
default_std = 1 / math.sqrt(hidden_dim) / 3.0
std_default = paddle.attr.Param(initial_std=default_std)
std_0 = paddle.attr.Param(initial_std=0.)
predicate_embedding = paddle.layer.embedding(
size=word_dim,
input=predicate,
param_attr=paddle.attr.Param(
name='vemb', initial_std=default_std))
mark_embedding = paddle.layer.embedding(
size=mark_dim, input=mark, param_attr=std_0)
word_input = [word, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2]
emb_layers = [
paddle.layer.embedding(
size=word_dim, input=x, param_attr=emb_para) for x in word_input
]
emb_layers.append(predicate_embedding)
emb_layers.append(mark_embedding)
```
3. 8个LSTM单元以“正向/反向”的顺序对所有输入序列进行学习。
```python
hidden_0 = paddle.layer.mixed(
- 3. 8个LSTM单元以“正向/反向”的顺序对所有输入序列进行学习。
```python
hidden_0 = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=emb, param_attr=std_default) for emb in emb_layers
])
mix_hidden_lr = 1e-3
lstm_para_attr = paddle.attr.Param(initial_std=0.0, learning_rate=1.0)
hidden_para_attr = paddle.attr.Param(
initial_std=default_std, learning_rate=mix_hidden_lr)
lstm_0 = paddle.layer.lstmemory(
input=hidden_0,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
bias_attr=std_0,
param_attr=lstm_para_attr)
#stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0]
for i in range(1, depth):
mix_hidden = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=emb, param_attr=std_default) for emb in emb_layers
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
])
mix_hidden_lr = 1e-3
lstm_para_attr = paddle.attr.Param(initial_std=0.0, learning_rate=1.0)
hidden_para_attr = paddle.attr.Param(
initial_std=default_std, learning_rate=mix_hidden_lr)
lstm_0 = paddle.layer.lstmemory(
input=hidden_0,
lstm = paddle.layer.lstmemory(
input=mix_hidden,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
reverse=((i % 2) == 1),
bias_attr=std_0,
param_attr=lstm_para_attr)
#stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0]
for i in range(1, depth):
mix_hidden = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
])
lstm = paddle.layer.lstmemory(
input=mix_hidden,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
reverse=((i % 2) == 1),
bias_attr=std_0,
param_attr=lstm_para_attr)
input_tmp = [mix_hidden, lstm]
```
4. 取最后一个栈式LSTM的输出和这个LSTM单元的输入到隐层映射，经过一个全连接层映射到标记字典的维度，得到最终的特征向量表示。
```python
feature_out = paddle.layer.mixed(
input_tmp = [mix_hidden, lstm]
```
- 4. 取最后一个栈式LSTM的输出和这个LSTM单元的输入到隐层映射，经过一个全连接层映射到标记字典的维度，得到最终的特征向量表示。
```python
feature_out = paddle.layer.mixed(
size=label_dict_len,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
], )
```
- 5. 网络的末端定义CRF层计算损失(cost)，指定参数名字为 `crfw`，该层需要输入正确的数据标签(target)。
```python
crf_cost = paddle.layer.crf(
size=label_dict_len,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
], )
```
5. 网络的末端定义CRF层计算损失(cost)，指定参数名字为 `crfw`，该层需要输入正确的数据标签(target)。
```python
crf_cost = paddle.layer.crf(
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(
name='crfw',
initial_std=default_std,
learning_rate=mix_hidden_lr))
```
6. CRF译码层和CRF层参数名字相同，即共享权重。如果输入了正确的数据标签(target)，会统计错误标签的个数，可以用来评估模型。如果没有输入正确的数据标签，该层可以推到出最优解，可以用来预测模型。
```python
crf_dec = paddle.layer.crf_decoding(
name='crf_dec_l',
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(name='crfw'))
```
input=feature_out,
label=target,
param_attr=paddle.attr.Param(
name='crfw',
initial_std=default_std,
learning_rate=mix_hidden_lr))
```
- 6. CRF译码层和CRF层参数名字相同，即共享权重。如果输入了正确的数据标签(target)，会统计错误标签的个数，可以用来评估模型。如果没有输入正确的数据标签，该层可以推到出最优解，可以用来预测模型。
```python
crf_dec = paddle.layer.crf_decoding(
name='crf_dec_l',
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(name='crfw'))
```
## 训练模型
......@@ -376,8 +380,8 @@ print parameters.keys()
# 这里加载PaddlePaddle上版保存的二进制模型
def load_parameter(file_name, h, w):
with open(file_name, 'rb') as f:
f.read(16)
return np.fromfile(f, dtype=np.float32).reshape(h, w)
f.read(16)
return np.fromfile(f, dtype=np.float32).reshape(h, w)
parameters.set('emb', load_parameter(conll05.get_embedding(), 44068, 32))
```
......
label_semantic_roles/index.en.html
浏览文件 @ 9e6a24cd
......@@ -266,6 +266,8 @@ We trained in the English Wikipedia language model to get a word vector lookup t
Get dictionary, print dictionary size:
```python
import math
import numpy as np
import paddle.v2 as paddle
import paddle.v2.dataset.conll05 as conll05
......@@ -274,164 +276,164 @@ word_dict_len = len(word_dict)
label_dict_len = len(label_dict)
pred_len = len(verb_dict)
print len(word_dict_len)
print len(label_dict_len)
print len(pred_len)
print word_dict_len
print label_dict_len
print pred_len
```
## Model configuration
1. Define input data dimensions and model hyperparameters.
- 1. Define input data dimensions and model hyperparameters.
```python
mark_dict_len = 2 # Value range of region mark. Region mark is either 0 or 1, so range is 2
word_dim = 32 # word vector dimension
mark_dim = 5 # adjacent dimension
hidden_dim = 512 # the dimension of LSTM hidden layer vector is 128 (512/4)
depth = 8 # depth of stacked LSTM
# There are 9 features per sample, so we will define 9 data layers.
# They type for each layer is integer_value_sequence.
def d_type(value_range):
return paddle.data_type.integer_value_sequence(value_range)
# word sequence
word = paddle.layer.data(name='word_data', type=d_type(word_dict_len))
# predicate
predicate = paddle.layer.data(name='verb_data', type=d_type(pred_len))
# 5 features for predicate context
ctx_n2 = paddle.layer.data(name='ctx_n2_data', type=d_type(word_dict_len))
ctx_n1 = paddle.layer.data(name='ctx_n1_data', type=d_type(word_dict_len))
ctx_0 = paddle.layer.data(name='ctx_0_data', type=d_type(word_dict_len))
ctx_p1 = paddle.layer.data(name='ctx_p1_data', type=d_type(word_dict_len))
ctx_p2 = paddle.layer.data(name='ctx_p2_data', type=d_type(word_dict_len))
# region marker sequence
mark = paddle.layer.data(name='mark_data', type=d_type(mark_dict_len))
# label sequence
target = paddle.layer.data(name='target', type=d_type(label_dict_len))
```
```python
mark_dict_len = 2 # Value range of region mark. Region mark is either 0 or 1, so range is 2
word_dim = 32 # word vector dimension
mark_dim = 5 # adjacent dimension
hidden_dim = 512 # the dimension of LSTM hidden layer vector is 128 (512/4)
depth = 8 # depth of stacked LSTM
# There are 9 features per sample, so we will define 9 data layers.
# They type for each layer is integer_value_sequence.
def d_type(value_range):
return paddle.data_type.integer_value_sequence(value_range)
# word sequence
word = paddle.layer.data(name='word_data', type=d_type(word_dict_len))
# predicate
predicate = paddle.layer.data(name='verb_data', type=d_type(pred_len))
# 5 features for predicate context
ctx_n2 = paddle.layer.data(name='ctx_n2_data', type=d_type(word_dict_len))
ctx_n1 = paddle.layer.data(name='ctx_n1_data', type=d_type(word_dict_len))
ctx_0 = paddle.layer.data(name='ctx_0_data', type=d_type(word_dict_len))
ctx_p1 = paddle.layer.data(name='ctx_p1_data', type=d_type(word_dict_len))
ctx_p2 = paddle.layer.data(name='ctx_p2_data', type=d_type(word_dict_len))
# region marker sequence
mark = paddle.layer.data(name='mark_data', type=d_type(mark_dict_len))
# label sequence
target = paddle.layer.data(name='target', type=d_type(label_dict_len))
```
Speciala note: hidden_dim = 512 means LSTM hidden vector of 128 dimension (512/4). Please refer PaddlePaddle official documentation for detail: [lstmemory](http://www.paddlepaddle.org/doc/ui/api/trainer_config_helpers/layers.html#lstmemory)。
Speciala note: hidden_dim = 512 means LSTM hidden vector of 128 dimension (512/4). Please refer PaddlePaddle official documentation for detail: [lstmemory](http://www.paddlepaddle.org/doc/ui/api/trainer_config_helpers/layers.html#lstmemory)。
- 2. The word sequence, predicate, predicate context, and region mark sequence are transformed into embedding vector sequences.
```python
# Since word vectorlookup table is pre-trained, we won't update it this time.
# is_static being True prevents updating the lookup table during training.
emb_para = paddle.attr.Param(name='emb', initial_std=0., is_static=True)
# hyperparameter configurations
default_std = 1 / math.sqrt(hidden_dim) / 3.0
std_default = paddle.attr.Param(initial_std=default_std)
std_0 = paddle.attr.Param(initial_std=0.)
predicate_embedding = paddle.layer.embedding(
size=word_dim,
input=predicate,
param_attr=paddle.attr.Param(
name='vemb', initial_std=default_std))
mark_embedding = paddle.layer.embedding(
size=mark_dim, input=mark, param_attr=std_0)
word_input = [word, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2]
emb_layers = [
paddle.layer.embedding(
size=word_dim, input=x, param_attr=emb_para) for x in word_input
]
emb_layers.append(predicate_embedding)
emb_layers.append(mark_embedding)
```
2. The word sequence, predicate, predicate context, and region mark sequence are transformed into embedding vector sequences.
- 3. 8 LSTM units will be trained in "forward / backward" order.
```python
# Since word vectorlookup table is pre-trained, we won't update it this time.
# is_static being True prevents updating the lookup table during training.
emb_para = paddle.attr.Param(name='emb', initial_std=0., is_static=True)
# hyperparameter configurations
default_std = 1 / math.sqrt(hidden_dim) / 3.0
std_default = paddle.attr.Param(initial_std=default_std)
std_0 = paddle.attr.Param(initial_std=0.)
predicate_embedding = paddle.layer.embedding(
size=word_dim,
input=predicate,
param_attr=paddle.attr.Param(
name='vemb', initial_std=default_std))
mark_embedding = paddle.layer.embedding(
size=mark_dim, input=mark, param_attr=std_0)
word_input = [word, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2]
emb_layers = [
paddle.layer.embedding(
size=word_dim, input=x, param_attr=emb_para) for x in word_input
]
emb_layers.append(predicate_embedding)
emb_layers.append(mark_embedding)
```
3. 8 LSTM units will be trained in "forward / backward" order.
```python
hidden_0 = paddle.layer.mixed(
```python
hidden_0 = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=emb, param_attr=std_default) for emb in emb_layers
])
mix_hidden_lr = 1e-3
lstm_para_attr = paddle.attr.Param(initial_std=0.0, learning_rate=1.0)
hidden_para_attr = paddle.attr.Param(
initial_std=default_std, learning_rate=mix_hidden_lr)
lstm_0 = paddle.layer.lstmemory(
input=hidden_0,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
bias_attr=std_0,
param_attr=lstm_para_attr)
# stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0]
for i in range(1, depth):
mix_hidden = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=emb, param_attr=std_default) for emb in emb_layers
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
])
mix_hidden_lr = 1e-3
lstm_para_attr = paddle.attr.Param(initial_std=0.0, learning_rate=1.0)
hidden_para_attr = paddle.attr.Param(
initial_std=default_std, learning_rate=mix_hidden_lr)
lstm_0 = paddle.layer.lstmemory(
input=hidden_0,
lstm = paddle.layer.lstmemory(
input=mix_hidden,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
reverse=((i % 2) == 1),
bias_attr=std_0,
param_attr=lstm_para_attr)
# stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0]
for i in range(1, depth):
mix_hidden = paddle.layer.mixed(
size=hidden_dim,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
])
lstm = paddle.layer.lstmemory(
input=mix_hidden,
act=paddle.activation.Relu(),
gate_act=paddle.activation.Sigmoid(),
state_act=paddle.activation.Sigmoid(),
reverse=((i % 2) == 1),
bias_attr=std_0,
param_attr=lstm_para_attr)
input_tmp = [mix_hidden, lstm]
```
4. We will concatenate the output of top LSTM unit with it's input, and project into a hidden layer. Then put a fully connected layer on top of it to get the final vector representation.
```python
feature_out = paddle.layer.mixed(
input_tmp = [mix_hidden, lstm]
```
- 4. We will concatenate the output of top LSTM unit with it's input, and project into a hidden layer. Then put a fully connected layer on top of it to get the final vector representation.
```python
feature_out = paddle.layer.mixed(
size=label_dict_len,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
], )
```
- 5. We use CRF as cost function, the parameter of CRF cost will be named `crfw`.
```python
crf_cost = paddle.layer.crf(
size=label_dict_len,
bias_attr=std_default,
input=[
paddle.layer.full_matrix_projection(
input=input_tmp[0], param_attr=hidden_para_attr),
paddle.layer.full_matrix_projection(
input=input_tmp[1], param_attr=lstm_para_attr)
], )
```
5. We use CRF as cost function, the parameter of CRF cost will be named `crfw`.
```python
crf_cost = paddle.layer.crf(
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(
name='crfw',
initial_std=default_std,
learning_rate=mix_hidden_lr))
```
6. CRF decoding layer is used for evaluation and inference. It shares parameter with CRF layer. The sharing of parameters among multiple layers is specified by the same parameter name in these layers.
```python
crf_dec = paddle.layer.crf_decoding(
name='crf_dec_l',
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(name='crfw'))
```
input=feature_out,
label=target,
param_attr=paddle.attr.Param(
name='crfw',
initial_std=default_std,
learning_rate=mix_hidden_lr))
```
- 6. CRF decoding layer is used for evaluation and inference. It shares parameter with CRF layer. The sharing of parameters among multiple layers is specified by the same parameter name in these layers.
```python
crf_dec = paddle.layer.crf_decoding(
name='crf_dec_l',
size=label_dict_len,
input=feature_out,
label=target,
param_attr=paddle.attr.Param(name='crfw'))
```
## Train model
......@@ -454,8 +456,8 @@ Now we load pre-trained word lookup table.
```python
def load_parameter(file_name, h, w):
with open(file_name, 'rb') as f:
f.read(16)
return np.fromfile(f, dtype=np.float32).reshape(h, w)
f.read(16)
return np.fromfile(f, dtype=np.float32).reshape(h, w)
parameters.set('emb', load_parameter(conll05.get_embedding(), 44068, 32))
```
......@@ -481,15 +483,15 @@ trainer = paddle.trainer.SGD(cost=crf_cost,
As mentioned in data preparation section, we will use CoNLL 2005 test corpus as training data set. `conll05.test()` outputs one training instance at a time. It will be shuffled, and batched into mini batches as input.
```python
reader = paddle.reader.batched(
reader = paddle.batch(
paddle.reader.shuffle(
conll05.test(), buf_size=8192), batch_size=20)
```
`reader_dict` is used to specify relationship between data instance and layer layer. For example, according to following `reader_dict`, the 0th column of data instance produced by`conll05.test()` correspond to data layer named `word_data`.
`feeding` is used to specify relationship between data instance and layer layer. For example, according to following `feeding`, the 0th column of data instance produced by`conll05.test()` correspond to data layer named `word_data`.
```python
reader_dict = {
feeding = {
'word_data': 0,
'ctx_n2_data': 1,
'ctx_n1_data': 2,
......@@ -519,7 +521,7 @@ trainer.train(
reader=reader,
event_handler=event_handler,
num_passes=10000,
reader_dict=reader_dict)
feeding=feeding)
```
## Conclusion
......
label_semantic_roles/index.html
浏览文件 @ 9e6a24cd
此差异已折叠。
label_semantic_roles/train.py
浏览文件 @ 9e6a24cd
......@@ -159,7 +159,7 @@ def main():
paddle.reader.shuffle(
conll05.test(), buf_size=8192), batch_size=10)
reader_dict = {
feeding = {
'word_data': 0,
'ctx_n2_data': 1,
'ctx_n1_data': 2,
......@@ -181,7 +181,7 @@ def main():
reader=reader,
event_handler=event_handler,
num_passes=10000,
reader_dict=reader_dict)
feeding=feeding)
if __name__ == '__main__':
......
recognize_digits/README.md
浏览文件 @ 9e6a24cd
此差异已折叠。
recognize_digits/index.html
浏览文件 @ 9e6a24cd
......@@ -108,16 +108,37 @@ Softmax回归模型采用了最简单的两层神经网络，即只有输入层
### 卷积神经网络(Convolutional Neural Network, CNN)
在多层感知器模型中，将图像展开成一维向量输入到网络中，忽略了图像的位置和结构信息，而卷积神经网络能够更好的利用图像的结构信息。[LeNet-5](http://yann.lecun.com/exdb/lenet/)是一个较简单的卷积神经网络。图6显示了其结构：输入的二维图像，先经过两次卷积层到池化层，再经过全连接层，最后使用softmax分类作为输出层。下面我们主要介绍卷积层和池化层。
<p align="center">
<img src="image/cnn.png"><br/>
图6. LeNet-5卷积神经网络结构<br/>
</p>
#### 卷积层
卷积层是卷积神经网络的核心基石。在图像识别里我们提到的卷积是二维卷积，即离散二维滤波器（也称作卷积核）与二维图像做卷积操作，简单的讲是二维滤波器滑动到二维图像上所有位置，并在每个位置上与该像素点及其领域像素点做内积。卷积操作被广泛应用与图像处理领域，不同卷积核可以提取不同的特征，例如边沿、线性、角等特征。在深层卷积神经网络中，通过卷积操作可以提取出图像低级到复杂的特征。
<p align="center">
<img src="image/conv_layer.png" width=500><br/>
<img src="image/conv_layer.png"><br/>
图4. 卷积层图片<br/>
</p>
卷积层是卷积神经网络的核心基石。该层的参数由一组可学习的过滤器（也叫作卷积核）组成。在前向过程中，每个卷积核在输入层进行横向和纵向的扫描，与输入层对应扫描位置进行卷积，得到的结果加上偏置并用相应的激活函数进行激活，结果能够得到一个二维的激活图(activation map)。每个特定的卷积核都能得到特定的激活图(activation map)，如有的卷积核可能对识别边角，有的可能识别圆圈，那这些卷积核可能对于对应的特征响应要强。
图4给出一个卷积计算过程的示例图，输入图像大小为$H=5,W=5,D=3$，即$5 \times 5$大小的3通道（RGB，也称作深度）彩色图像。这个示例图中包含两（用$K$表示）组卷积核，即图中滤波器$W_0$和$W_1$。在卷积计算中，通常对不同的输入通道采用不同的卷积核，如图示例中每组卷积核包含（$D=3）$个$3 \times 3$（用$F \times F$表示）大小的卷积核。另外，这个示例中卷积核在图像的水平方向（$W$方向）和垂直方向（$H$方向）的滑动步长为2（用$S$表示）；对输入图像周围各填充1（用$P$表示）个0，即图中输入层原始数据为蓝色部分，灰色部分是进行了大小为1的扩展，用0来进行扩展。经过卷积操作得到输出为$3 \times 3 \times 2$（用$H_{o} \times W_{o} \times K$表示）大小的特征图，即$3 \times 3$大小的2通道特征图，其中$H_o$计算公式为：$H_o = (H - F + 2 \times P)/S + 1$，$W_o$同理。 而输出特征图中的每个像素，是每组滤波器与输入图像每个特征图的内积再求和，再加上偏置$b_o$，偏置通常对于每个输出特征图是共享的。例如图中输出特征图$o[:,:,0]$中的第一个$2$计算如下：
$$ o[0,0,0] = \sum x[0:3,0:3,0] * w_{0}[:,:,0]] + \sum x[0:3,0:3,1] * w_{0}[:,:,1]] + \sum x[0:3,0:3,2] * w_{0}[:,:,2]] + b_0 = 2 $$
$$ \sum x[0:3,0:3,0] * w_{0}[:,:,0]] = 0*1 + 0*1 + 0*1 + 0*1 + 1*1 + 2*(-1) + 0*(-1) + 0*1 + 0*(-1) = -1 $$
$$ \sum x[0:3,0:3,1] * w_{0}[:,:,1]] = 0*0 + 0*1 + 0*1 + 0*(-1) + 0*0 + 1*1 + 0*1 + 2*0 + 1*1 = 2 $$
$$ \sum x[0:3,0:3,2] * w_{0}[:,:,2]] = 0*(-1) + 0*1 + 0*(-1) + 0*0 + 1*1 + 1*0 + 0*(-1) + 1*0 + 1*(-1) = 0 $$
$$ b_0 = 1 $$
在卷积操作中卷积核是可学习的参数，经过上面示例介绍，每层卷积的参数大小为$D \times F \times F \times K$。在多层感知器模型中，神经元通常是全部连接，参数较多。而卷积层的参数较少，这也是由卷积层的主要特性即局部连接和共享权重所决定。
- 局部连接：每个神经元仅与输入神经元的一块区域连接，这块局部区域称作感受野（receptive field）。在图像卷积操作中，即神经元在空间维度（spatial dimension，即上图示例H和W所在的平面）是局部连接，但在深度上是全部连接。对于二维图像本身而言，也是局部像素关联较强。这种局部连接保证了学习后的过滤器能够对于局部的输入特征有最强的响应。局部连接的思想，也是受启发于生物学里面的视觉系统结构，视觉皮层的神经元就是局部接受信息的。
图4是卷积层的一个动态图。由于3D量难以表示，所有的3D量（输入的3D量（蓝色），权重3D量（红色），输出3D量（绿色））通过将深度在行上堆叠来表示。如图4，输入层是$W_1=5,H_1=5,D_1=3$，我们常见的彩色图片其实就是类似这样的输入层，彩色图片的宽和高对应这里的$W_1$和$H_1$，而彩色图片有RGB三个颜色通道，对应这里的$D_1$；卷积层的参数为$K=2,F=3,S=2,P=1$，这里的$K$是卷积核的数量，如图4中有$Filter W_0$和$Filter W_1$两个卷积核，$F$对应卷积核的大小，图中$W0$和$W1$在每一层深度上都是$3\times3$的矩阵，$S$对应卷积核扫描的步长，从动态图中可以看到，方框每次左移或下移2个单位，$P$对应Padding扩展，是对输入层的扩展，图中输入层，原始数据为蓝色部分，可以看到灰色部分是进行了大小为1的扩展，用0来进行扩展；图4的动态可视化对输出层结果（绿色）进行迭代，显示每个输出元素是通过将突出显示的输入（蓝色）与滤波器（红色）进行元素相乘，将其相加，然后通过偏置抵消结果来计算的。
- 权重共享：计算同一个深度切片的神经元时采用的滤波器是共享的。例如图4中计算$o[:,:,0]$的每个每个神经元的滤波器均相同，都为$W_0$，这样可以很大程度上减少参数。共享权重在一定程度上讲是有意义的，例如图片的底层边缘特征与特征在图中的具体位置无关。但是在一些场景中是无意的，比如输入的图片是人脸，眼睛和头发位于不同的位置，希望在不同的位置学到不同的特征 (参考[斯坦福大学公开课]( http://cs231n.github.io/convolutional-networks/))。请注意权重只是对于同一深度切片的神经元是共享的，在卷积层，通常采用多组卷积核提取不同特征，即对应不同深度切片的特征，不同深度切片的神经元权重是不共享。另外，偏重对同一深度切片的所有神经元都是共享的。
通过介绍卷积计算过程及其特性，可以看出卷积是线性操作，并具有平移不变性（shift-invariant），平移不变性即在图像每个位置执行相同的操作。卷积层的局部连接和权重共享使得需要学习的参数大大减小，这样也有利于训练较大卷积神经网络。
#### 池化层
......@@ -128,19 +149,6 @@ Softmax回归模型采用了最简单的两层神经网络，即只有输入层
池化是非线性下采样的一种形式，主要作用是通过减少网络的参数来减小计算量，并且能够在一定程度上控制过拟合。通常在卷积层的后面会加上一个池化层。池化包括最大池化、平均池化等。其中最大池化是用不重叠的矩形框将输入层分成不同的区域，对于每个矩形框的数取最大值作为输出层，如图5所示。
#### LeNet-5网络
<p align="center">
<img src="image/cnn.png"><br/>
图6. LeNet-5卷积神经网络结构<br/>
</p>
[LeNet-5](http://yann.lecun.com/exdb/lenet/)是一个最简单的卷积神经网络。图6显示了其结构：输入的二维图像，先经过两次卷积层到池化层，再经过全连接层，最后使用softmax分类作为输出层。卷积的如下三个特性，决定了LeNet-5能比同样使用全连接层的多层感知器更好地识别图像：
- 神经元的三维特性： 卷积层的神经元在宽度、高度和深度上进行了组织排列。每一层的神经元仅仅与前一层的一块小区域连接，这块小区域被称为感受野(receptive field)。
- 局部连接：CNN通过在相邻层的神经元之间实施局部连接模式来利用空间局部相关性。这样的结构保证了学习后的过滤器能够对于局部的输入特征有最强的响应。堆叠许多这样的层导致非线性“过滤器”变得越来越“全局”。这允许网络首先创建输入的小部分的良好表示，然后从它们组合较大区域的表示。
- 共享权重：在CNN中，每个滤波器在整个视野中重复扫描。 这些复制单元共享相同的参数化（权重向量和偏差）并形成特征图。 这意味着给定卷积层中的所有神经元检测完全相同的特征。 以这种方式的复制单元允许不管它们在视野中的位置都能检测到特征，从而构成平移不变性的性质。
更详细的关于卷积神经网络的具体知识可以参考[斯坦福大学公开课]( http://cs231n.github.io/convolutional-networks/ )和[图像分类](https://github.com/PaddlePaddle/book/blob/develop/image_classification/README.md)教程。
### 常见激活函数介绍
......
understand_sentiment/index.html
浏览文件 @ 9e6a24cd
此差异已折叠。
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