Skip to content

B
book

PaddlePaddle / book

通知 16
Star 4
Fork 0
B
book
未验证 提交 54ace206 编写于 作者: N Nicky Chan 提交者： GitHub
浏览文件
操作

Merge pull request #539 from sidgoyal78/modify_sent 

Modify sentiment analysis book chapter
上级 b326f9fe 89d558cc
显示空白变更内容
内联 并排
Showing with 873 addition and 471 deletion
06.understand_sentiment/README.md
浏览文件 @ 54ace206
...@@ -98,18 +98,21 @@ We use [IMDB](http://ai.stanford.edu/%7Eamaas/data/sentiment/) dataset for senti ...@@ -98,18 +98,21 @@ We use [IMDB](http://ai.stanford.edu/%7Eamaas/data/sentiment/) dataset for senti
After issuing a command `python train.py`, training will start immediately. The details will be unpacked by the following sessions to see how it works. After issuing a command `python train.py`, training will start immediately. The details will be unpacked by the following sessions to see how it works.
## Model Structure ## Model Configuration
### Initialize PaddlePaddle Our program starts with importing necessary packages and initializing some global variables:
We must import and initialize PaddlePaddle (enable/disable GPU, set the number of trainers, etc).
```python ```python
import sys import paddle
import paddle.v2 as paddle import paddle.fluid as fluid
from functools import partial
# PaddlePaddle init import numpy as np
paddle.init(use_gpu=False, trainer_count=1)
CLASS_DIM = 2
EMB_DIM = 128
HID_DIM = 512
BATCH_SIZE = 128
USE_GPU = False
``` ```
As alluded to in section [Model Overview](#model-overview), here we provide the implementations of both Text CNN and Stacked-bidirectional LSTM models. As alluded to in section [Model Overview](#model-overview), here we provide the implementations of both Text CNN and Stacked-bidirectional LSTM models.
...@@ -118,212 +121,229 @@ As alluded to in section [Model Overview](#model-overview), here we provide the ...@@ -118,212 +121,229 @@ As alluded to in section [Model Overview](#model-overview), here we provide the
We create a neural network `convolution_net` as the following snippet code. We create a neural network `convolution_net` as the following snippet code.
Note: `paddle.networks.sequence_conv_pool` includes both convolution and pooling layer operations. Note: `fluid.nets.sequence_conv_pool` includes both convolution and pooling layer operations.
```python ```python
def convolution_net(input_dim, class_dim=2, emb_dim=128, hid_dim=128): def convolution_net(data, input_dim, class_dim, emb_dim, hid_dim):
data = paddle.layer.data("word", emb = fluid.layers.embedding(
paddle.data_type.integer_value_sequence(input_dim)) input=data, size=[input_dim, emb_dim], is_sparse=True)
emb = paddle.layer.embedding(input=data, size=emb_dim) conv_3 = fluid.nets.sequence_conv_pool(
conv_3 = paddle.networks.sequence_conv_pool( input=emb,
input=emb, context_len=3, hidden_size=hid_dim) num_filters=hid_dim,
conv_4 = paddle.networks.sequence_conv_pool( filter_size=3,
input=emb, context_len=4, hidden_size=hid_dim) act="tanh",
output = paddle.layer.fc(input=[conv_3, conv_4], pool_type="sqrt")
size=class_dim, conv_4 = fluid.nets.sequence_conv_pool(
act=paddle.activation.Softmax()) input=emb,
lbl = paddle.layer.data("label", paddle.data_type.integer_value(2)) num_filters=hid_dim,
cost = paddle.layer.classification_cost(input=output, label=lbl) filter_size=4,
return cost, output act="tanh",
``` pool_type="sqrt")
prediction = fluid.layers.fc(
input=[conv_3, conv_4], size=class_dim, act="softmax")
return prediction
1. Define input data and its dimension ```
Parameter `input_dim` denotes the dictionary size, and `class_dim` is the number of categories.
Parameter `input_dim` denotes the dictionary size, and `class_dim` is the number of categories. In `convolution_net`, the input to the network is defined in `paddle.layer.data`.
1. Define Classifier
The above Text CNN network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category.
1. Define Loss Function The above Text CNN network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category.
In the context of supervised learning, labels of the training set are defined in `paddle.layer.data`, too. During training, cross-entropy is used as loss function in `paddle.layer.classification_cost` and as the output of the network; During testing, the outputs are the probabilities calculated in the classifier.
#### Stacked bidirectional LSTM ### Stacked bidirectional LSTM
We create a neural network `stacked_lstm_net` as below. We create a neural network `stacked_lstm_net` as below.
```python ```python
def stacked_lstm_net(input_dim, def stacked_lstm_net(data, input_dim, class_dim, emb_dim, hid_dim, stacked_num):
class_dim=2,
emb_dim=128, emb = fluid.layers.embedding(
hid_dim=512, input=data, size=[input_dim, emb_dim], is_sparse=True)
stacked_num=3):
""" fc1 = fluid.layers.fc(input=emb, size=hid_dim)
A Wrapper for sentiment classification task. lstm1, cell1 = fluid.layers.dynamic_lstm(input=fc1, size=hid_dim)
This network uses a bi-directional recurrent network,
consisting of three LSTM layers. This configuration is
motivated from the following paper, but uses few layers.
http://www.aclweb.org/anthology/P15-1109
input_dim: here is word dictionary dimension.
class_dim: number of categories.
emb_dim: dimension of word embedding.
hid_dim: dimension of hidden layer.
stacked_num: number of stacked lstm-hidden layer.
"""
assert stacked_num % 2 == 1
fc_para_attr = paddle.attr.Param(learning_rate=1e-3)
lstm_para_attr = paddle.attr.Param(initial_std=0., learning_rate=1.)
para_attr = [fc_para_attr, lstm_para_attr]
bias_attr = paddle.attr.Param(initial_std=0., l2_rate=0.)
relu = paddle.activation.Relu()
linear = paddle.activation.Linear()
data = paddle.layer.data("word",
paddle.data_type.integer_value_sequence(input_dim))
emb = paddle.layer.embedding(input=data, size=emb_dim)
fc1 = paddle.layer.fc(input=emb,
size=hid_dim,
act=linear,
bias_attr=bias_attr)
lstm1 = paddle.layer.lstmemory(
input=fc1, act=relu, bias_attr=bias_attr)
inputs = [fc1, lstm1] inputs = [fc1, lstm1]
for i in range(2, stacked_num + 1): for i in range(2, stacked_num + 1):
fc = paddle.layer.fc(input=inputs, fc = fluid.layers.fc(input=inputs, size=hid_dim)
size=hid_dim, lstm, cell = fluid.layers.dynamic_lstm(
act=linear, input=fc, size=hid_dim, is_reverse=(i % 2) == 0)
param_attr=para_attr,
bias_attr=bias_attr)
lstm = paddle.layer.lstmemory(
input=fc,
reverse=(i % 2) == 0,
act=relu,
bias_attr=bias_attr)
inputs = [fc, lstm] inputs = [fc, lstm]
fc_last = paddle.layer.pooling( fc_last = fluid.layers.sequence_pool(input=inputs[0], pool_type='max')
input=inputs[0], pooling_type=paddle.pooling.Max()) lstm_last = fluid.layers.sequence_pool(input=inputs[1], pool_type='max')
lstm_last = paddle.layer.pooling(
input=inputs[1], pooling_type=paddle.pooling.Max()) prediction = fluid.layers.fc(input=[fc_last, lstm_last],
output = paddle.layer.fc(input=[fc_last, lstm_last],
size=class_dim, size=class_dim,
act=paddle.activation.Softmax(), act='softmax')
bias_attr=bias_attr, return prediction
param_attr=para_attr)
lbl = paddle.layer.data("label", paddle.data_type.integer_value(2))
cost = paddle.layer.classification_cost(input=output, label=lbl)
return cost, output
``` ```
The above stacked bidirectional LSTM network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category.
1. Define input data and its dimension To reiterate, we can either invoke `convolution_net` or `stacked_lstm_net`. In below steps, we will go with `convolution_net`.
Parameter `input_dim` denotes the dictionary size, and `class_dim` is the number of categories. In `stacked_lstm_net`, the input to the network is defined in `paddle.layer.data`. Next we define an `inference_program` that simply uses `convolution_net` to predict output with the input from `fluid.layer.data`.
1. Define Classifier ```python
def inference_program(word_dict):
The above stacked bidirectional LSTM network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category. data = fluid.layers.data(
name="words", shape=[1], dtype="int64", lod_level=1)
1. Define Loss Function dict_dim = len(word_dict)
net = convolution_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM)
return net
```
In the context of supervised learning, labels of the training set are defined in `paddle.layer.data`, too. During training, cross-entropy is used as loss function in `paddle.layer.classification_cost` and as the output of the network; During testing, the outputs are the probabilities calculated in the classifier. Then we define a `training_program` that uses the result from `inference_program` to compute the cost with label data.
Also define `optimizer_func` to specify the optimizer.
To reiterate, we can either invoke `convolution_net` or `stacked_lstm_net`. In the context of supervised learning, labels of the training set are defined in `paddle.layer.data` too. During training, cross-entropy is used as loss function in `paddle.layer.classification_cost` and as the output of the network; During testing, the outputs are the probabilities calculated in the classifier.
First result that returns from the list must be cost.
```python ```python
word_dict = paddle.dataset.imdb.word_dict() def train_program(word_dict):
dict_dim = len(word_dict) prediction = inference_program(word_dict)
class_dim = 2 label = fluid.layers.data(name="label", shape=[1], dtype="int64")
cost = fluid.layers.cross_entropy(input=prediction, label=label)
avg_cost = fluid.layers.mean(cost)
accuracy = fluid.layers.accuracy(input=prediction, label=label)
return [avg_cost, accuracy]
# option 1 def optimizer_func():
[cost, output] = convolution_net(dict_dim, class_dim=class_dim) return fluid.optimizer.Adagrad(learning_rate=0.002)
# option 2
# [cost, output] = stacked_lstm_net(dict_dim, class_dim=class_dim, stacked_num=3)
``` ```
## Model Training ## Model Training
### Define Parameters ### Specify training environment
First, we create the model parameters according to the previous model configuration `cost`. Specify your training environment, you should specify if the training is on CPU or GPU.
```python ```python
# create parameters use_cuda = False
parameters = paddle.parameters.create(cost) place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
``` ```
### Create Trainer ### Datafeeder Configuration
Before jumping into creating a training module, algorithm setting is also necessary. Next we define data feeders for test and train. The feeder reads a `buf_size` of data each time and feed them to the training/testing process.
Here we specified `Adam` optimization algorithm via `paddle.optimizer`. `paddle.dataset.imdb.train` will yield records during each pass, after shuffling, a batch input of `BATCH_SIZE` is generated for training.
```python Notice for loading and reading IMDB data, it could take up to 1 minute. Please be patient.
# create optimizer
adam_optimizer = paddle.optimizer.Adam(
learning_rate=2e-3,
regularization=paddle.optimizer.L2Regularization(rate=8e-4),
model_average=paddle.optimizer.ModelAverage(average_window=0.5))
# create trainer
trainer = paddle.trainer.SGD(cost=cost,
parameters=parameters,
update_equation=adam_optimizer)
```
### Training ```python
`paddle.dataset.imdb.train()` will yield records during each pass, after shuffling, a batch input is generated for training. print("Loading IMDB word dict....")
word_dict = paddle.dataset.imdb.word_dict()
```python print ("Reading training data....")
train_reader = paddle.batch( train_reader = paddle.batch(
paddle.reader.shuffle( paddle.reader.shuffle(
lambda: paddle.dataset.imdb.train(word_dict), buf_size=1000), paddle.dataset.imdb.train(word_dict), buf_size=25000),
batch_size=100) batch_size=BATCH_SIZE)
```
test_reader = paddle.batch(
lambda: paddle.dataset.imdb.test(word_dict), batch_size=100) ### Create Trainer
Create a trainer that takes `train_program` as input and specify optimizer function.
```python
trainer = fluid.Trainer(
train_func=partial(train_program, word_dict),
place=place,
optimizer_func=optimizer_func)
``` ```
`feeding` is devoted to specifying the correspondence between each yield record and `paddle.layer.data`. For instance, the first column of data generated by `paddle.dataset.imdb.train()` corresponds to `word` feature. ### Feeding Data
`feed_order` is devoted to specifying the correspondence between each yield record and `paddle.layer.data`. For instance, the first column of data generated by `imdb.train` corresponds to `words`.
```python ```python
feeding = {'word': 0, 'label': 1} feed_order = ['words', 'label']
``` ```
Callback function `event_handler` will be invoked to track training progress when a pre-defined event happens. ### Event Handler
Callback function `event_handler` will be called during training when a pre-defined event happens.
For example, we can check the cost by `trainer.test` when `EndStepEvent` occurs
```python ```python
# Specify the directory path to save the parameters
params_dirname = "understand_sentiment_conv.inference.model"
def event_handler(event): def event_handler(event):
if isinstance(event, paddle.event.EndIteration): if isinstance(event, fluid.EndStepEvent):
if event.batch_id % 100 == 0: print("Step {0}, Epoch {1} Metrics {2}".format(
print "\nPass %d, Batch %d, Cost %f, %s" % ( event.step, event.epoch, map(np.array, event.metrics)))
event.pass_id, event.batch_id, event.cost, event.metrics)
else: if event.step == 10:
sys.stdout.write('.') trainer.save_params(params_dirname)
sys.stdout.flush() trainer.stop()
if isinstance(event, paddle.event.EndPass):
with open('./params_pass_%d.tar' % event.pass_id, 'w') as f:
trainer.save_parameter_to_tar(f)
result = trainer.test(reader=test_reader, feeding=feeding)
print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
``` ```
Finally, we can invoke `trainer.train` to start training: ### Training
Finally, we invoke `trainer.train` to start training with `num_epochs` and other parameters.
```python ```python
trainer.train( trainer.train(
reader=train_reader, num_epochs=1,
event_handler=event_handler, event_handler=event_handler,
feeding=feeding, reader=train_reader,
num_passes=10) feed_order=feed_order)
``` ```
## Inference
### Create Inferencer
Initialize Inferencer with `inference_program` and `params_dirname` which is where we save params from training.
```python
inferencer = fluid.Inferencer(
infer_func=partial(inference_program, word_dict),
param_path=params_dirname,
place=place)
```
### Create Lod Tensor with test data
To do inference, we pick 3 potential reviews out of our mind as testing data. Feel free to modify any of them.
We map each word in the reviews to id from `word_dict`, replaced by 'unknown' if the word is not in `word_dict`.
Then we create lod data with the id list and use `create_lod_tensor` to create lod tensor.
```python
reviews_str = [
'read the book forget the movie', 'this is a great movie', 'this is very bad'
]
reviews = [c.split() for c in reviews_str]
UNK = word_dict['<unk>']
lod = []
for c in reviews:
lod.append([word_dict.get(words, UNK) for words in c])
base_shape = [[len(c) for c in lod]]
tensor_words = fluid.create_lod_tensor(lod, base_shape, place)
```
### Infer
Now we can infer and predict probability of positive or negative from each review above.
```python
results = inferencer.infer({'words': tensor_words})
for i, r in enumerate(results[0]):
print("Predict probability of ", r[0], " to be positive and ", r[1], " to be negative for review \'", reviews_str[i], "\'")
```
## Conclusion ## Conclusion
......
06.understand_sentiment/index.html
浏览文件 @ 54ace206
...@@ -140,18 +140,21 @@ We use [IMDB](http://ai.stanford.edu/%7Eamaas/data/sentiment/) dataset for senti ...@@ -140,18 +140,21 @@ We use [IMDB](http://ai.stanford.edu/%7Eamaas/data/sentiment/) dataset for senti
After issuing a command `python train.py`, training will start immediately. The details will be unpacked by the following sessions to see how it works. After issuing a command `python train.py`, training will start immediately. The details will be unpacked by the following sessions to see how it works.
## Model Structure ## Model Configuration
### Initialize PaddlePaddle Our program starts with importing necessary packages and initializing some global variables:
We must import and initialize PaddlePaddle (enable/disable GPU, set the number of trainers, etc).
```python ```python
import sys import paddle
import paddle.v2 as paddle import paddle.fluid as fluid
from functools import partial
# PaddlePaddle init import numpy as np
paddle.init(use_gpu=False, trainer_count=1)
CLASS_DIM = 2
EMB_DIM = 128
HID_DIM = 512
BATCH_SIZE = 128
USE_GPU = False
``` ```
As alluded to in section [Model Overview](#model-overview), here we provide the implementations of both Text CNN and Stacked-bidirectional LSTM models. As alluded to in section [Model Overview](#model-overview), here we provide the implementations of both Text CNN and Stacked-bidirectional LSTM models.
...@@ -160,212 +163,229 @@ As alluded to in section [Model Overview](#model-overview), here we provide the ...@@ -160,212 +163,229 @@ As alluded to in section [Model Overview](#model-overview), here we provide the
We create a neural network `convolution_net` as the following snippet code. We create a neural network `convolution_net` as the following snippet code.
Note: `paddle.networks.sequence_conv_pool` includes both convolution and pooling layer operations. Note: `fluid.nets.sequence_conv_pool` includes both convolution and pooling layer operations.
```python ```python
def convolution_net(input_dim, class_dim=2, emb_dim=128, hid_dim=128): def convolution_net(data, input_dim, class_dim, emb_dim, hid_dim):
data = paddle.layer.data("word", emb = fluid.layers.embedding(
paddle.data_type.integer_value_sequence(input_dim)) input=data, size=[input_dim, emb_dim], is_sparse=True)
emb = paddle.layer.embedding(input=data, size=emb_dim) conv_3 = fluid.nets.sequence_conv_pool(
conv_3 = paddle.networks.sequence_conv_pool( input=emb,
input=emb, context_len=3, hidden_size=hid_dim) num_filters=hid_dim,
conv_4 = paddle.networks.sequence_conv_pool( filter_size=3,
input=emb, context_len=4, hidden_size=hid_dim) act="tanh",
output = paddle.layer.fc(input=[conv_3, conv_4], pool_type="sqrt")
size=class_dim, conv_4 = fluid.nets.sequence_conv_pool(
act=paddle.activation.Softmax()) input=emb,
lbl = paddle.layer.data("label", paddle.data_type.integer_value(2)) num_filters=hid_dim,
cost = paddle.layer.classification_cost(input=output, label=lbl) filter_size=4,
return cost, output act="tanh",
``` pool_type="sqrt")
prediction = fluid.layers.fc(
input=[conv_3, conv_4], size=class_dim, act="softmax")
return prediction
1. Define input data and its dimension ```
Parameter `input_dim` denotes the dictionary size, and `class_dim` is the number of categories.
Parameter `input_dim` denotes the dictionary size, and `class_dim` is the number of categories. In `convolution_net`, the input to the network is defined in `paddle.layer.data`.
1. Define Classifier
The above Text CNN network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category.
1. Define Loss Function The above Text CNN network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category.
In the context of supervised learning, labels of the training set are defined in `paddle.layer.data`, too. During training, cross-entropy is used as loss function in `paddle.layer.classification_cost` and as the output of the network; During testing, the outputs are the probabilities calculated in the classifier.
#### Stacked bidirectional LSTM ### Stacked bidirectional LSTM
We create a neural network `stacked_lstm_net` as below. We create a neural network `stacked_lstm_net` as below.
```python ```python
def stacked_lstm_net(input_dim, def stacked_lstm_net(data, input_dim, class_dim, emb_dim, hid_dim, stacked_num):
class_dim=2,
emb_dim=128, emb = fluid.layers.embedding(
hid_dim=512, input=data, size=[input_dim, emb_dim], is_sparse=True)
stacked_num=3):
""" fc1 = fluid.layers.fc(input=emb, size=hid_dim)
A Wrapper for sentiment classification task. lstm1, cell1 = fluid.layers.dynamic_lstm(input=fc1, size=hid_dim)
This network uses a bi-directional recurrent network,
consisting of three LSTM layers. This configuration is
motivated from the following paper, but uses few layers.
http://www.aclweb.org/anthology/P15-1109
input_dim: here is word dictionary dimension.
class_dim: number of categories.
emb_dim: dimension of word embedding.
hid_dim: dimension of hidden layer.
stacked_num: number of stacked lstm-hidden layer.
"""
assert stacked_num % 2 == 1
fc_para_attr = paddle.attr.Param(learning_rate=1e-3)
lstm_para_attr = paddle.attr.Param(initial_std=0., learning_rate=1.)
para_attr = [fc_para_attr, lstm_para_attr]
bias_attr = paddle.attr.Param(initial_std=0., l2_rate=0.)
relu = paddle.activation.Relu()
linear = paddle.activation.Linear()
data = paddle.layer.data("word",
paddle.data_type.integer_value_sequence(input_dim))
emb = paddle.layer.embedding(input=data, size=emb_dim)
fc1 = paddle.layer.fc(input=emb,
size=hid_dim,
act=linear,
bias_attr=bias_attr)
lstm1 = paddle.layer.lstmemory(
input=fc1, act=relu, bias_attr=bias_attr)
inputs = [fc1, lstm1] inputs = [fc1, lstm1]
for i in range(2, stacked_num + 1): for i in range(2, stacked_num + 1):
fc = paddle.layer.fc(input=inputs, fc = fluid.layers.fc(input=inputs, size=hid_dim)
size=hid_dim, lstm, cell = fluid.layers.dynamic_lstm(
act=linear, input=fc, size=hid_dim, is_reverse=(i % 2) == 0)
param_attr=para_attr,
bias_attr=bias_attr)
lstm = paddle.layer.lstmemory(
input=fc,
reverse=(i % 2) == 0,
act=relu,
bias_attr=bias_attr)
inputs = [fc, lstm] inputs = [fc, lstm]
fc_last = paddle.layer.pooling( fc_last = fluid.layers.sequence_pool(input=inputs[0], pool_type='max')
input=inputs[0], pooling_type=paddle.pooling.Max()) lstm_last = fluid.layers.sequence_pool(input=inputs[1], pool_type='max')
lstm_last = paddle.layer.pooling(
input=inputs[1], pooling_type=paddle.pooling.Max()) prediction = fluid.layers.fc(input=[fc_last, lstm_last],
output = paddle.layer.fc(input=[fc_last, lstm_last],
size=class_dim, size=class_dim,
act=paddle.activation.Softmax(), act='softmax')
bias_attr=bias_attr, return prediction
param_attr=para_attr)
lbl = paddle.layer.data("label", paddle.data_type.integer_value(2))
cost = paddle.layer.classification_cost(input=output, label=lbl)
return cost, output
``` ```
The above stacked bidirectional LSTM network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category.
1. Define input data and its dimension To reiterate, we can either invoke `convolution_net` or `stacked_lstm_net`. In below steps, we will go with `convolution_net`.
Parameter `input_dim` denotes the dictionary size, and `class_dim` is the number of categories. In `stacked_lstm_net`, the input to the network is defined in `paddle.layer.data`. Next we define an `inference_program` that simply uses `convolution_net` to predict output with the input from `fluid.layer.data`.
1. Define Classifier ```python
def inference_program(word_dict):
The above stacked bidirectional LSTM network extracts high-level features and maps them to a vector of the same size as the categories. `paddle.activation.Softmax` function or classifier is then used for calculating the probability of the sentence belonging to each category. data = fluid.layers.data(
name="words", shape=[1], dtype="int64", lod_level=1)
1. Define Loss Function dict_dim = len(word_dict)
net = convolution_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM)
return net
```
In the context of supervised learning, labels of the training set are defined in `paddle.layer.data`, too. During training, cross-entropy is used as loss function in `paddle.layer.classification_cost` and as the output of the network; During testing, the outputs are the probabilities calculated in the classifier. Then we define a `training_program` that uses the result from `inference_program` to compute the cost with label data.
Also define `optimizer_func` to specify the optimizer.
To reiterate, we can either invoke `convolution_net` or `stacked_lstm_net`. In the context of supervised learning, labels of the training set are defined in `paddle.layer.data` too. During training, cross-entropy is used as loss function in `paddle.layer.classification_cost` and as the output of the network; During testing, the outputs are the probabilities calculated in the classifier.
First result that returns from the list must be cost.
```python ```python
word_dict = paddle.dataset.imdb.word_dict() def train_program(word_dict):
dict_dim = len(word_dict) prediction = inference_program(word_dict)
class_dim = 2 label = fluid.layers.data(name="label", shape=[1], dtype="int64")
cost = fluid.layers.cross_entropy(input=prediction, label=label)
avg_cost = fluid.layers.mean(cost)
accuracy = fluid.layers.accuracy(input=prediction, label=label)
return [avg_cost, accuracy]
# option 1 def optimizer_func():
[cost, output] = convolution_net(dict_dim, class_dim=class_dim) return fluid.optimizer.Adagrad(learning_rate=0.002)
# option 2
# [cost, output] = stacked_lstm_net(dict_dim, class_dim=class_dim, stacked_num=3)
``` ```
## Model Training ## Model Training
### Define Parameters ### Specify training environment
First, we create the model parameters according to the previous model configuration `cost`. Specify your training environment, you should specify if the training is on CPU or GPU.
```python ```python
# create parameters use_cuda = False
parameters = paddle.parameters.create(cost) place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
``` ```
### Create Trainer ### Datafeeder Configuration
Before jumping into creating a training module, algorithm setting is also necessary. Next we define data feeders for test and train. The feeder reads a `buf_size` of data each time and feed them to the training/testing process.
Here we specified `Adam` optimization algorithm via `paddle.optimizer`. `paddle.dataset.imdb.train` will yield records during each pass, after shuffling, a batch input of `BATCH_SIZE` is generated for training.
```python Notice for loading and reading IMDB data, it could take up to 1 minute. Please be patient.
# create optimizer
adam_optimizer = paddle.optimizer.Adam(
learning_rate=2e-3,
regularization=paddle.optimizer.L2Regularization(rate=8e-4),
model_average=paddle.optimizer.ModelAverage(average_window=0.5))
# create trainer
trainer = paddle.trainer.SGD(cost=cost,
parameters=parameters,
update_equation=adam_optimizer)
```
### Training ```python
`paddle.dataset.imdb.train()` will yield records during each pass, after shuffling, a batch input is generated for training. print("Loading IMDB word dict....")
word_dict = paddle.dataset.imdb.word_dict()
```python print ("Reading training data....")
train_reader = paddle.batch( train_reader = paddle.batch(
paddle.reader.shuffle( paddle.reader.shuffle(
lambda: paddle.dataset.imdb.train(word_dict), buf_size=1000), paddle.dataset.imdb.train(word_dict), buf_size=25000),
batch_size=100) batch_size=BATCH_SIZE)
```
test_reader = paddle.batch(
lambda: paddle.dataset.imdb.test(word_dict), batch_size=100) ### Create Trainer
Create a trainer that takes `train_program` as input and specify optimizer function.
```python
trainer = fluid.Trainer(
train_func=partial(train_program, word_dict),
place=place,
optimizer_func=optimizer_func)
``` ```
`feeding` is devoted to specifying the correspondence between each yield record and `paddle.layer.data`. For instance, the first column of data generated by `paddle.dataset.imdb.train()` corresponds to `word` feature. ### Feeding Data
`feed_order` is devoted to specifying the correspondence between each yield record and `paddle.layer.data`. For instance, the first column of data generated by `imdb.train` corresponds to `words`.
```python ```python
feeding = {'word': 0, 'label': 1} feed_order = ['words', 'label']
``` ```
Callback function `event_handler` will be invoked to track training progress when a pre-defined event happens. ### Event Handler
Callback function `event_handler` will be called during training when a pre-defined event happens.
For example, we can check the cost by `trainer.test` when `EndStepEvent` occurs
```python ```python
# Specify the directory path to save the parameters
params_dirname = "understand_sentiment_conv.inference.model"
def event_handler(event): def event_handler(event):
if isinstance(event, paddle.event.EndIteration): if isinstance(event, fluid.EndStepEvent):
if event.batch_id % 100 == 0: print("Step {0}, Epoch {1} Metrics {2}".format(
print "\nPass %d, Batch %d, Cost %f, %s" % ( event.step, event.epoch, map(np.array, event.metrics)))
event.pass_id, event.batch_id, event.cost, event.metrics)
else: if event.step == 10:
sys.stdout.write('.') trainer.save_params(params_dirname)
sys.stdout.flush() trainer.stop()
if isinstance(event, paddle.event.EndPass):
with open('./params_pass_%d.tar' % event.pass_id, 'w') as f:
trainer.save_parameter_to_tar(f)
result = trainer.test(reader=test_reader, feeding=feeding)
print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
``` ```
Finally, we can invoke `trainer.train` to start training: ### Training
Finally, we invoke `trainer.train` to start training with `num_epochs` and other parameters.
```python ```python
trainer.train( trainer.train(
reader=train_reader, num_epochs=1,
event_handler=event_handler, event_handler=event_handler,
feeding=feeding, reader=train_reader,
num_passes=10) feed_order=feed_order)
``` ```
## Inference
### Create Inferencer
Initialize Inferencer with `inference_program` and `params_dirname` which is where we save params from training.
```python
inferencer = fluid.Inferencer(
infer_func=partial(inference_program, word_dict),
param_path=params_dirname,
place=place)
```
### Create Lod Tensor with test data
To do inference, we pick 3 potential reviews out of our mind as testing data. Feel free to modify any of them.
We map each word in the reviews to id from `word_dict`, replaced by 'unknown' if the word is not in `word_dict`.
Then we create lod data with the id list and use `create_lod_tensor` to create lod tensor.
```python
reviews_str = [
'read the book forget the movie', 'this is a great movie', 'this is very bad'
]
reviews = [c.split() for c in reviews_str]
UNK = word_dict['<unk>']
lod = []
for c in reviews:
lod.append([word_dict.get(words, UNK) for words in c])
base_shape = [[len(c) for c in lod]]
tensor_words = fluid.create_lod_tensor(lod, base_shape, place)
```
### Infer
Now we can infer and predict probability of positive or negative from each review above.
```python
results = inferencer.infer({'words': tensor_words})
for i, r in enumerate(results[0]):
print("Predict probability of ", r[0], " to be positive and ", r[1], " to be negative for review \'", reviews_str[i], "\'")
```
## Conclusion ## Conclusion
......
06.understand_sentiment/train.py 已删除 100644 → 0
浏览文件 @ b326f9fe
# 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, os
import paddle.v2 as paddle
with_gpu = os.getenv('WITH_GPU', '0') != '0'
def convolution_net(input_dim, class_dim=2, emb_dim=128, hid_dim=128):
data = paddle.layer.data("word",
paddle.data_type.integer_value_sequence(input_dim))
emb = paddle.layer.embedding(input=data, size=emb_dim)
conv_3 = paddle.networks.sequence_conv_pool(
input=emb, context_len=3, hidden_size=hid_dim)
conv_4 = paddle.networks.sequence_conv_pool(
input=emb, context_len=4, hidden_size=hid_dim)
output = paddle.layer.fc(
input=[conv_3, conv_4], size=class_dim, act=paddle.activation.Softmax())
lbl = paddle.layer.data("label", paddle.data_type.integer_value(2))
cost = paddle.layer.classification_cost(input=output, label=lbl)
return cost, output
def stacked_lstm_net(input_dim,
class_dim=2,
emb_dim=128,
hid_dim=512,
stacked_num=3):
"""
A Wrapper for sentiment classification task.
This network uses bi-directional recurrent network,
consisting three LSTM layers. This configure is referred to
the paper as following url, but use fewer layrs.
http://www.aclweb.org/anthology/P15-1109
input_dim: here is word dictionary dimension.
class_dim: number of categories.
emb_dim: dimension of word embedding.
hid_dim: dimension of hidden layer.
stacked_num: number of stacked lstm-hidden layer.
"""
assert stacked_num % 2 == 1
fc_para_attr = paddle.attr.Param(learning_rate=1e-3)
lstm_para_attr = paddle.attr.Param(initial_std=0., learning_rate=1.)
para_attr = [fc_para_attr, lstm_para_attr]
bias_attr = paddle.attr.Param(initial_std=0., l2_rate=0.)
relu = paddle.activation.Relu()
linear = paddle.activation.Linear()
data = paddle.layer.data("word",
paddle.data_type.integer_value_sequence(input_dim))
emb = paddle.layer.embedding(input=data, size=emb_dim)
fc1 = paddle.layer.fc(
input=emb, size=hid_dim, act=linear, bias_attr=bias_attr)
lstm1 = paddle.layer.lstmemory(input=fc1, act=relu, bias_attr=bias_attr)
inputs = [fc1, lstm1]
for i in range(2, stacked_num + 1):
fc = paddle.layer.fc(
input=inputs,
size=hid_dim,
act=linear,
param_attr=para_attr,
bias_attr=bias_attr)
lstm = paddle.layer.lstmemory(
input=fc, reverse=(i % 2) == 0, act=relu, bias_attr=bias_attr)
inputs = [fc, lstm]
fc_last = paddle.layer.pooling(
input=inputs[0], pooling_type=paddle.pooling.Max())
lstm_last = paddle.layer.pooling(
input=inputs[1], pooling_type=paddle.pooling.Max())
output = paddle.layer.fc(
input=[fc_last, lstm_last],
size=class_dim,
act=paddle.activation.Softmax(),
bias_attr=bias_attr,
param_attr=para_attr)
lbl = paddle.layer.data("label", paddle.data_type.integer_value(2))
cost = paddle.layer.classification_cost(input=output, label=lbl)
return cost, output
if __name__ == '__main__':
# init
paddle.init(use_gpu=with_gpu)
#data
print 'load dictionary...'
word_dict = paddle.dataset.imdb.word_dict()
dict_dim = len(word_dict)
class_dim = 2
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.imdb.train(word_dict), buf_size=1000),
batch_size=100)
test_reader = paddle.batch(
paddle.dataset.imdb.test(word_dict), batch_size=100)
feeding = {'word': 0, 'label': 1}
# network config
# Please choose the way to build the network
# by uncommenting the corresponding line.
[cost, output] = convolution_net(dict_dim, class_dim=class_dim)
# [cost, output] = stacked_lstm_net(dict_dim, class_dim=class_dim, stacked_num=3)
# create parameters
parameters = paddle.parameters.create(cost)
# create optimizer
adam_optimizer = paddle.optimizer.Adam(
learning_rate=2e-3,
regularization=paddle.optimizer.L2Regularization(rate=8e-4),
model_average=paddle.optimizer.ModelAverage(average_window=0.5))
# create trainer
trainer = paddle.trainer.SGD(
cost=cost, parameters=parameters, update_equation=adam_optimizer)
# 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):
with open('./params_pass_%d.tar' % event.pass_id, 'w') as f:
trainer.save_parameter_to_tar(f)
result = trainer.test(reader=test_reader, feeding=feeding)
print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
# Save the inference topology to protobuf.
inference_topology = paddle.topology.Topology(layers=output)
with open("./inference_topology.pkl", 'wb') as f:
inference_topology.serialize_for_inference(f)
trainer.train(
reader=train_reader,
event_handler=event_handler,
feeding=feeding,
num_passes=20)
06.understand_sentiment/train_conv.py 0 → 100644
浏览文件 @ 54ace206
# Copyright (c) 2018 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.
from __future__ import print_function
import os
import paddle
import paddle.fluid as fluid
from functools import partial
import numpy as np
CLASS_DIM = 2
EMB_DIM = 128
HID_DIM = 512
BATCH_SIZE = 128
def convolution_net(data, input_dim, class_dim, emb_dim, hid_dim):
emb = fluid.layers.embedding(
input=data, size=[input_dim, emb_dim], is_sparse=True)
conv_3 = fluid.nets.sequence_conv_pool(
input=emb,
num_filters=hid_dim,
filter_size=3,
act="tanh",
pool_type="sqrt")
conv_4 = fluid.nets.sequence_conv_pool(
input=emb,
num_filters=hid_dim,
filter_size=4,
act="tanh",
pool_type="sqrt")
prediction = fluid.layers.fc(
input=[conv_3, conv_4], size=class_dim, act="softmax")
return prediction
def inference_program(word_dict):
data = fluid.layers.data(
name="words", shape=[1], dtype="int64", lod_level=1)
dict_dim = len(word_dict)
net = convolution_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM)
return net
def train_program(word_dict):
prediction = inference_program(word_dict)
label = fluid.layers.data(name="label", shape=[1], dtype="int64")
cost = fluid.layers.cross_entropy(input=prediction, label=label)
avg_cost = fluid.layers.mean(cost)
accuracy = fluid.layers.accuracy(input=prediction, label=label)
return [avg_cost, accuracy]
def optimizer_func():
return fluid.optimizer.Adagrad(learning_rate=0.002)
def train(use_cuda, train_program, params_dirname):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
print("Loading IMDB word dict....")
word_dict = paddle.dataset.imdb.word_dict()
print("Reading training data....")
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.imdb.train(word_dict), buf_size=25000),
batch_size=BATCH_SIZE)
print("Reading testing data....")
test_reader = paddle.batch(
paddle.dataset.imdb.test(word_dict), batch_size=BATCH_SIZE)
trainer = fluid.Trainer(
train_func=partial(train_program, word_dict),
place=place,
optimizer_func=optimizer_func)
feed_order = ['words', 'label']
def event_handler(event):
if isinstance(event, fluid.EndStepEvent):
if event.step % 10 == 0:
avg_cost, acc = trainer.test(
reader=test_reader, feed_order=feed_order)
print('Step {0}, Test Loss {1:0.2}, Acc {2:0.2}'.format(
event.step, avg_cost, acc))
print("Step {0}, Epoch {1} Metrics {2}".format(
event.step, event.epoch, map(np.array, event.metrics)))
elif isinstance(event, fluid.EndEpochEvent):
trainer.save_params(params_dirname)
trainer.train(
num_epochs=1,
event_handler=event_handler,
reader=train_reader,
feed_order=feed_order)
def infer(use_cuda, inference_program, params_dirname=None):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
word_dict = paddle.dataset.imdb.word_dict()
inferencer = fluid.Inferencer(
infer_func=partial(inference_program, word_dict),
param_path=params_dirname,
place=place)
# Setup input by creating LoDTensor to represent sequence of words.
# Here each word is the basic element of the LoDTensor and the shape of
# each word (base_shape) should be [1] since it is simply an index to
# look up for the corresponding word vector.
# Suppose the length_based level of detail (lod) info is set to [[3, 4, 2]],
# which has only one lod level. Then the created LoDTensor will have only
# one higher level structure (sequence of words, or sentence) than the basic
# element (word). Hence the LoDTensor will hold data for three sentences of
# length 3, 4 and 2, respectively.
# Note that lod info should be a list of lists.
reviews_str = [
'read the book forget the movie', 'this is a great movie',
'this is very bad'
]
reviews = [c.split() for c in reviews_str]
UNK = word_dict['<unk>']
lod = []
for c in reviews:
lod.append([word_dict.get(words, UNK) for words in c])
base_shape = [[len(c) for c in lod]]
tensor_words = fluid.create_lod_tensor(lod, base_shape, place)
results = inferencer.infer({'words': tensor_words})
for i, r in enumerate(results[0]):
print("Predict probability of ", r[0], " to be positive and ", r[1],
" to be negative for review \'", reviews_str[i], "\'")
def main(use_cuda):
if use_cuda and not fluid.core.is_compiled_with_cuda():
return
params_dirname = "understand_sentiment_conv.inference.model"
train(use_cuda, train_program, params_dirname)
infer(use_cuda, inference_program, params_dirname)
if __name__ == '__main__':
use_cuda = os.getenv('WITH_GPU', '0') != '0'
main(use_cuda)
06.understand_sentiment/train_dyn_rnn.py 0 → 100644
浏览文件 @ 54ace206
# Copyright (c) 2018 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.
from __future__ import print_function
import os
import paddle
import paddle.fluid as fluid
from functools import partial
import numpy as np
CLASS_DIM = 2
EMB_DIM = 128
BATCH_SIZE = 128
LSTM_SIZE = 128
USE_GPU = False
def dynamic_rnn_lstm(data, input_dim, class_dim, emb_dim, lstm_size):
emb = fluid.layers.embedding(
input=data, size=[input_dim, emb_dim], is_sparse=True)
sentence = fluid.layers.fc(input=emb, size=lstm_size, act='tanh')
rnn = fluid.layers.DynamicRNN()
with rnn.block():
word = rnn.step_input(sentence)
prev_hidden = rnn.memory(value=0.0, shape=[lstm_size])
prev_cell = rnn.memory(value=0.0, shape=[lstm_size])
def gate_common(ipt, hidden, size):
gate0 = fluid.layers.fc(input=ipt, size=size, bias_attr=True)
gate1 = fluid.layers.fc(input=hidden, size=size, bias_attr=False)
return gate0 + gate1
forget_gate = fluid.layers.sigmoid(x=gate_common(word, prev_hidden,
lstm_size))
input_gate = fluid.layers.sigmoid(x=gate_common(word, prev_hidden,
lstm_size))
output_gate = fluid.layers.sigmoid(x=gate_common(word, prev_hidden,
lstm_size))
cell_gate = fluid.layers.sigmoid(x=gate_common(word, prev_hidden,
lstm_size))
cell = forget_gate * prev_cell + input_gate * cell_gate
hidden = output_gate * fluid.layers.tanh(x=cell)
rnn.update_memory(prev_cell, cell)
rnn.update_memory(prev_hidden, hidden)
rnn.output(hidden)
last = fluid.layers.sequence_last_step(rnn())
prediction = fluid.layers.fc(input=last, size=class_dim, act="softmax")
return prediction
def inference_program(word_dict):
data = fluid.layers.data(
name="words", shape=[1], dtype="int64", lod_level=1)
dict_dim = len(word_dict)
pred = dynamic_rnn_lstm(data, dict_dim, CLASS_DIM, EMB_DIM, LSTM_SIZE)
return pred
def train_program(word_dict):
prediction = inference_program(word_dict)
label = fluid.layers.data(name="label", shape=[1], dtype="int64")
cost = fluid.layers.cross_entropy(input=prediction, label=label)
avg_cost = fluid.layers.mean(cost)
accuracy = fluid.layers.accuracy(input=prediction, label=label)
return [avg_cost, accuracy]
def optimizer_func():
return fluid.optimizer.Adagrad(learning_rate=0.002)
def train(use_cuda, train_program, params_dirname):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
print("Loading IMDB word dict....")
word_dict = paddle.dataset.imdb.word_dict()
print("Reading training data....")
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.imdb.train(word_dict), buf_size=25000),
batch_size=BATCH_SIZE)
print("Reading testing data....")
test_reader = paddle.batch(
paddle.dataset.imdb.test(word_dict), batch_size=BATCH_SIZE)
trainer = fluid.Trainer(
train_func=partial(train_program, word_dict),
place=place,
optimizer_func=optimizer_func)
feed_order = ['words', 'label']
def event_handler(event):
if isinstance(event, fluid.EndStepEvent):
if event.step % 10 == 0:
avg_cost, acc = trainer.test(
reader=test_reader, feed_order=feed_order)
print('Step {0}, Test Loss {1:0.2}, Acc {2:0.2}'.format(
event.step, avg_cost, acc))
print("Step {0}, Epoch {1} Metrics {2}".format(
event.step, event.epoch, map(np.array, event.metrics)))
elif isinstance(event, fluid.EndEpochEvent):
trainer.save_params(params_dirname)
trainer.train(
num_epochs=1,
event_handler=event_handler,
reader=train_reader,
feed_order=feed_order)
def infer(use_cuda, inference_program, params_dirname=None):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
word_dict = paddle.dataset.imdb.word_dict()
inferencer = fluid.Inferencer(
infer_func=partial(inference_program, word_dict),
param_path=params_dirname,
place=place)
# Setup input by creating LoDTensor to represent sequence of words.
# Here each word is the basic element of the LoDTensor and the shape of
# each word (base_shape) should be [1] since it is simply an index to
# look up for the corresponding word vector.
# Suppose the length_based level of detail (lod) info is set to [[3, 4, 2]],
# which has only one lod level. Then the created LoDTensor will have only
# one higher level structure (sequence of words, or sentence) than the basic
# element (word). Hence the LoDTensor will hold data for three sentences of
# length 3, 4 and 2, respectively.
# Note that lod info should be a list of lists.
reviews_str = [
'read the book forget the movie', 'this is a great movie',
'this is very bad'
]
reviews = [c.split() for c in reviews_str]
UNK = word_dict['<unk>']
lod = []
for c in reviews:
lod.append([word_dict.get(words, UNK) for words in c])
base_shape = [[len(c) for c in lod]]
tensor_words = fluid.create_lod_tensor(lod, base_shape, place)
results = inferencer.infer({'words': tensor_words})
for i, r in enumerate(results[0]):
print("Predict probability of ", r[0], " to be positive and ", r[1],
" to be negative for review \'", reviews_str[i], "\'")
def main(use_cuda):
if use_cuda and not fluid.core.is_compiled_with_cuda():
return
params_dirname = "understand_sentiment_conv.inference.model"
train(use_cuda, train_program, params_dirname)
infer(use_cuda, inference_program, params_dirname)
if __name__ == '__main__':
use_cuda = os.getenv('WITH_GPU', '0') != '0'
main(use_cuda)
06.understand_sentiment/train_stacked_lstm.py 0 → 100644
浏览文件 @ 54ace206
# Copyright (c) 2018 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.
from __future__ import print_function
import os
import paddle
import paddle.fluid as fluid
from functools import partial
import numpy as np
CLASS_DIM = 2
EMB_DIM = 128
HID_DIM = 512
STACKED_NUM = 3
BATCH_SIZE = 128
USE_GPU = False
def stacked_lstm_net(data, input_dim, class_dim, emb_dim, hid_dim, stacked_num):
assert stacked_num % 2 == 1
emb = fluid.layers.embedding(
input=data, size=[input_dim, emb_dim], is_sparse=True)
fc1 = fluid.layers.fc(input=emb, size=hid_dim)
lstm1, cell1 = fluid.layers.dynamic_lstm(input=fc1, size=hid_dim)
inputs = [fc1, lstm1]
for i in range(2, stacked_num + 1):
fc = fluid.layers.fc(input=inputs, size=hid_dim)
lstm, cell = fluid.layers.dynamic_lstm(
input=fc, size=hid_dim, is_reverse=(i % 2) == 0)
inputs = [fc, lstm]
fc_last = fluid.layers.sequence_pool(input=inputs[0], pool_type='max')
lstm_last = fluid.layers.sequence_pool(input=inputs[1], pool_type='max')
prediction = fluid.layers.fc(
input=[fc_last, lstm_last], size=class_dim, act='softmax')
return prediction
def inference_program(word_dict):
data = fluid.layers.data(
name="words", shape=[1], dtype="int64", lod_level=1)
dict_dim = len(word_dict)
net = stacked_lstm_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM,
STACKED_NUM)
return net
def train_program(word_dict):
prediction = inference_program(word_dict)
label = fluid.layers.data(name="label", shape=[1], dtype="int64")
cost = fluid.layers.cross_entropy(input=prediction, label=label)
avg_cost = fluid.layers.mean(cost)
accuracy = fluid.layers.accuracy(input=prediction, label=label)
return [avg_cost, accuracy]
def optimizer_func():
return fluid.optimizer.Adagrad(learning_rate=0.002)
def train(use_cuda, train_program, params_dirname):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
print("Loading IMDB word dict....")
word_dict = paddle.dataset.imdb.word_dict()
print("Reading training data....")
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.imdb.train(word_dict), buf_size=25000),
batch_size=BATCH_SIZE)
print("Reading testing data....")
test_reader = paddle.batch(
paddle.dataset.imdb.test(word_dict), batch_size=BATCH_SIZE)
trainer = fluid.Trainer(
train_func=partial(train_program, word_dict),
place=place,
optimizer_func=optimizer_func)
feed_order = ['words', 'label']
def event_handler(event):
if isinstance(event, fluid.EndStepEvent):
if event.step % 10 == 0:
avg_cost, acc = trainer.test(
reader=test_reader, feed_order=feed_order)
print('Step {0}, Test Loss {1:0.2}, Acc {2:0.2}'.format(
event.step, avg_cost, acc))
print("Step {0}, Epoch {1} Metrics {2}".format(
event.step, event.epoch, map(np.array, event.metrics)))
elif isinstance(event, fluid.EndEpochEvent):
trainer.save_params(params_dirname)
trainer.train(
num_epochs=1,
event_handler=event_handler,
reader=train_reader,
feed_order=feed_order)
def infer(use_cuda, inference_program, params_dirname=None):
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
word_dict = paddle.dataset.imdb.word_dict()
inferencer = fluid.Inferencer(
infer_func=partial(inference_program, word_dict),
param_path=params_dirname,
place=place)
# Setup input by creating LoDTensor to represent sequence of words.
# Here each word is the basic element of the LoDTensor and the shape of
# each word (base_shape) should be [1] since it is simply an index to
# look up for the corresponding word vector.
# Suppose the length_based level of detail (lod) info is set to [[3, 4, 2]],
# which has only one lod level. Then the created LoDTensor will have only
# one higher level structure (sequence of words, or sentence) than the basic
# element (word). Hence the LoDTensor will hold data for three sentences of
# length 3, 4 and 2, respectively.
# Note that lod info should be a list of lists.
reviews_str = [
'read the book forget the movie', 'this is a great movie',
'this is very bad'
]
reviews = [c.split() for c in reviews_str]
UNK = word_dict['<unk>']
lod = []
for c in reviews:
lod.append([word_dict.get(words, UNK) for words in c])
base_shape = [[len(c) for c in lod]]
tensor_words = fluid.create_lod_tensor(lod, base_shape, place)
results = inferencer.infer({'words': tensor_words})
for i, r in enumerate(results[0]):
print("Predict probability of ", r[0], " to be positive and ", r[1],
" to be negative for review \'", reviews_str[i], "\'")
def main(use_cuda):
if use_cuda and not fluid.core.is_compiled_with_cuda():
return
params_dirname = "understand_sentiment_stacked_lstm.inference.model"
train(use_cuda, train_program, params_dirname)
infer(use_cuda, inference_program, params_dirname)
if __name__ == '__main__':
use_cuda = os.getenv('WITH_GPU', '0') != '0'
main(use_cuda)
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