Merge pull request #2 from denglelaibh/denglelaibh

Realization of DRN 2016 (also similar to resNet 16)

drn/.idea/drn.iml

+<?xml version="1.0" encoding="UTF-8"?>
+<module type="PYTHON_MODULE" version="4">
+  <component name="NewModuleRootManager">
+    <content url="file://$MODULE_DIR$" />
+    <orderEntry type="inheritedJdk" />
+    <orderEntry type="sourceFolder" forTests="false" />
+  </component>
+  <component name="TestRunnerService">
+    <option name="projectConfiguration" value="Nosetests" />
+    <option name="PROJECT_TEST_RUNNER" value="Nosetests" />
+  </component>
+</module>
\ No newline at end of file

drn/.idea/misc.xml

+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="ProjectRootManager" version="2" project-jdk-name="Python 2.7.13 (C:\Users\67938\Anaconda2\python.exe)" project-jdk-type="Python SDK" />
+</project>
\ No newline at end of file

drn/.idea/modules.xml

+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="ProjectModuleManager">
+    <modules>
+      <module fileurl="file://$PROJECT_DIR$/.idea/drn.iml" filepath="$PROJECT_DIR$/.idea/drn.iml" />
+    </modules>
+  </component>
+</project>
\ No newline at end of file

drn/README.md

+The minimum PaddlePaddle version needed for the code sample in this directory is v0.10.0. If you are on a version of PaddlePaddle earlier than v0.10.0, please [update your installation](http://www.paddlepaddle.org/docs/develop/documentation/en/build_and_install/pip_install_en.html).
+-----------------------
+# Deep Residual Networks（DRN）
+## 简介
+在论文[1]中提到了，1202层的ResNet出现了过拟合的问题，有待进一步改进。第二年，何的团队就发表了“Identity Mappings in Deep Residual Networks”这篇文章[2]，分析了ResNet成功的关键因素——residual block背后的算法，并对residual block以及after-addition activation进行改进，通过一系列的ablation experiments验证了，在residual block和after-addition activation上都使用identity mapping（恒等映射）时，能对模型训练产生很好的效果，通过这项改进，也成功的训练出了具有很好效果的ResNet-1001。
+## DRN 网络结构
+在原始的ResNet中，对于每一个residual building block：
+![pic1](./img/pic1.png)
+
+可以表现为以下形式:   
+
+$$
+y_l = h(x_l) + f(x_l, w_l)
+x_{l+1} = f(y_l)
+$$
+
+其中$h(x_1)$为一个恒等映射，$f(y_l)$代表ReLU激活函数，在[2]中提出了，如果如果$h(x)$和$f(y)$都是恒等映射，即$h(x_l)=x_l、f(y_l)=y_l$，那么在训练的前向和反向传播阶段，信号可以直接从一个单元传递到另外一个单元，使得训练变得更加简单。 
+在DNN16中，具有以下优良特性：
+
+（1）对于任意深的单元**L**的特征 $x_L$ 可以表达为浅层单元**l**的特征 $x_l$加上一个形如 $\sum_{i=l}^{L−1}F$的残差函数，这表明了任意单元**L** 和 **l**之间都具有残差特性。 
+
+（2）对于任意深的单元 **L**，它的特征 $x_L = x_0 + \sum_{i=0}^{L−1}F(x_i,W_i)$，即为之前所有残差函数输出的总和(加上$x_0$)。而正好相反的是，“plain network”中的特征xL是一系列矩阵向量的乘积，也就是$\prod_{i=0}^{L−1}W_i x_0$，而求和的计算量远远小于求积的计算量。
+
+实验发现，$h(x_l) = x_l$的误差衰减最快，误差也最低(下图a子图所示)：  
+
+![pic2](./img/pic2.png)
+
+
+对于激活函数，验发现，将ReLU和BN都放在预激活中，即full pre-activation(下图子图e所示)在ResNet-110和ResNet-164上的效果都最好。
+
+![pic3](./img/pic3.png)
+
+## 复现文件一览
+在复现文件中，包含以下文件：
+
+<table>
+  <tr>
+    <th width=60%>文件</th>
+    <th width=40%>描述</th>
+  </tr>
+  <tr>
+    <td> train.py  </td>
+    <td> DRN模型训练脚本 </td>
+  </tr>
+  <tr>
+    <td> infer.py </td>
+    <td> 利用训练好的DRN模型做预测 </td>
+  <tr>
+    <td> drn.py </td>
+    <td> 定义DRN的网络结构 </td>
+  </tr>
+</table>
+
+## 基于flower数据集的模型复现
+### 数据准备
+所使用的的数据集是paddle中自带的flowers数据集进行训练，直接import即可:
+
+```
+import paddle.v2.dataset.flowers as flowers
+```
+
+### 网络定义
+网络的定义在文件```drn.py```中完整实现，其中最主要的是残差网络的部分：
+```
+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)
+
+```
+### 训练
+接下来，执行``` python train.py -model drn``` 即可训练过程，在训练过程中，建议使CUDA GPU进行训练，如果使用CPU训练耗时可长达90小时以上，关键代码为：
+
+```
+paddle.init(use_gpu=True, trainer_count=1)
+
+image = paddle.layer.data(name="image", type=paddle.data_type.dense_vector(DATA_DIM))
+
+lbl = paddle.layer.data(name="label", type=paddle.data_type.integer_value(CLASS_DIM))
+
+（省略部分代码）
+
+trainer = paddle.trainer.SGD(cost=cost,
+                             parameters=parameters,
+                             update_equation=optimizer,
+                             extra_layers=extra_layers)
+（省略部分代码）
+   
+trainer.train(
+        reader=train_reader, num_passes=200, event_handler=event_handler)
+
+```
+
+下面是关于上述代码的解释：
+
+1. 进行``` paddle.init ```以1个GPU的方式初始化
+
+2. 定义```img```图像名 和 ```lbl``` 图像标签
+ 
+3. 定义```trainer```，包含损失函数、参数、优化器和层数信息
+ 
+4. 在```train```函数中进行实际训练，共执行200趟
+
+执行过程中，控制台将打印如下所示的信息：
+```
+Pass 0, Batch 0, Cost 2.2512, ...
+Pass 0, Batch 1, Cost 2.1532, ...
+```
+
+同时在```train.py```目录下，每趟训练完成时，将生成```params_pass_0.tar,gz```，最后一趟的200.tar.gz文件生成时，训练完成。
+
+### 应用模型
+应用训练好的模型，执行``` python infer.py -data_list <文件目录> =model drn```即可：
+
+```
+
+\# load parameters
+with gzip.open('params_pass_200.tar.gz', 'r') as f:
+    parameters = paddle.parameters.Parameters.from_tar(f)
+
+file_list = [line.strip() for line in open(image_list_file)]
+test_data = [(paddle.image.load_and_transform(image_file, 256, 224, False)
+              .flatten().astype('float32'), )
+             for image_file in file_list]
+probs = paddle.infer(
+    output_layer=out, parameters=parameters, input=test_data)
+lab = np.argsort(-probs)
+for file_name, result in zip(file_list, lab):
+    print "Label of %s is: %d" % (file_name, result[0])
+
+```
+
+代码将从图片文件夹中读取对应的图片文件，同时给出预测的标签结果，并进行输出。

drn/drn.py

+import paddle.v2 as paddle
+
+__all__ = ['drn16']
+
+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(input, ch_out, stride):
+    if input.num_filters != ch_out:
+        return conv_bn_layer(input, ch_out, 1, stride, 0,
+                             paddle.activation.Linear())
+    else:
+        return input
+
+
+def basicblock(input, ch_out, stride):
+    short = shortcut(input, ch_out, stride)
+    conv1 = conv_bn_layer(input, ch_out, 3, stride, 1)
+    conv2 = conv_bn_layer(conv1, ch_out, 3, 1, 1, paddle.activation.Linear())
+    return paddle.layer.addto(
+        input=[short, conv2], act=paddle.activation.Relu())
+
+
+def bottleneck(input, ch_out, stride):
+    short = shortcut(input, ch_out * 4, stride)
+    conv1 = conv_bn_layer(input, ch_out, 1, stride, 0)
+    conv2 = conv_bn_layer(conv1, ch_out, 3, 1, 1)
+    conv3 = conv_bn_layer(conv2, ch_out * 4, 1, 1, 0,
+                          paddle.activation.Linear())
+    return paddle.layer.addto(
+        input=[short, conv3], act=paddle.activation.Relu())
+
+
+def layer_warp(block_func, input, ch_out, count, stride):
+    conv = block_func(input, ch_out, stride)
+    for i in range(1, count):
+        conv = block_func(conv, ch_out, 1)
+    return conv
+
+
+def drn16(input, class_dim, depth=32):
+    assert (depth - 2) % 6 == 0
+    n = (depth - 2) / 6
+    conv1 = conv_bn_layer(
+        input, 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())
+    out = paddle.layer.fc(input=pool,
+                          size=class_dim,
+                          act=paddle.activation.Softmax())
+    return out
\ No newline at end of file

drn/infer.py

+import os
+import gzip
+import argparse
+import numpy as np
+from PIL import Image
+
+import paddle.v2 as paddle
+import drn
+
+
+DATA_DIM = 3 * 224 * 224
+CLASS_DIM = 102
+
+
+def main():
+    # parse the argument
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        'data_list',
+        help='The path of data list file, which consists of one image path per line'
+    )
+    parser.add_argument(
+        'model',
+        help='The model for image classification',
+        choices=[
+            'drn'
+        ])
+    parser.add_argument(
+        'params_path', help='The file which stores the parameters')
+    args = parser.parse_args()
+
+    # PaddlePaddle init
+    paddle.init(use_gpu=True, trainer_count=1)
+
+    image = paddle.layer.data(
+        name="image", type=paddle.data_type.dense_vector(DATA_DIM))
+
+    if args.model == 'drn':
+        out = drn.drn16(image, class_dim=CLASS_DIM)
+
+    # load parameters
+    with gzip.open(args.params_path, 'r') as f:
+        parameters = paddle.parameters.Parameters.from_tar(f)
+
+    file_list = [line.strip() for line in open(args.data_list)]
+    test_data = [(paddle.image.load_and_transform(image_file, 256, 224, False)
+                  .flatten().astype('float32'), ) for image_file in file_list]
+    probs = paddle.infer(
+        output_layer=out, parameters=parameters, input=test_data)
+    lab = np.argsort(-probs)
+    for file_name, result in zip(file_list, lab):
+        print "Label of %s is: %d" % (file_name, result[0])
+
+
+if __name__ == '__main__':
+    main()
\ No newline at end of file

drn/train.py

+import gzip
+import argparse
+
+import paddle.v2.dataset.flowers as flowers
+import paddle.v2 as paddle
+import drn
+
+DATA_DIM = 3 * 224 * 224
+CLASS_DIM = 102
+BATCH_SIZE = 128
+
+
+def main():
+    # parse the argument
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        'model',
+        help='The model for image classification',
+        choices=[
+           'drn'
+        ])
+    args = parser.parse_args()
+
+    # PaddlePaddle init
+    paddle.init(use_gpu=True, trainer_count=1)
+
+    image = paddle.layer.data(
+        name="image", type=paddle.data_type.dense_vector(DATA_DIM))
+    lbl = paddle.layer.data(
+        name="label", type=paddle.data_type.integer_value(CLASS_DIM))
+
+    extra_layers = None
+    learning_rate = 0.01
+    if args.model == 'drn':
+        out = drn.drn16(image, class_dim=CLASS_DIM)
+
+    cost = paddle.layer.classification_cost(input=out, label=lbl)
+
+    # Create parameters
+    parameters = paddle.parameters.create(cost)
+
+    # Create optimizer
+    optimizer = paddle.optimizer.Momentum(
+        momentum=0.9,
+        regularization=paddle.optimizer.L2Regularization(rate=0.0005 *
+                                                         BATCH_SIZE),
+        learning_rate=learning_rate / BATCH_SIZE,
+        learning_rate_decay_a=0.1,
+        learning_rate_decay_b=128000 * 35,
+        learning_rate_schedule="discexp", )
+
+    train_reader = paddle.batch(
+        paddle.reader.shuffle(
+            flowers.train(),
+            # To use other data, replace the above line with:
+            # reader.train_reader('train.list'),
+            buf_size=1000),
+        batch_size=BATCH_SIZE)
+    test_reader = paddle.batch(
+        flowers.valid(),
+        # To use other data, replace the above line with:
+        # reader.test_reader('val.list'),
+        batch_size=BATCH_SIZE)
+
+    # Create trainer
+    trainer = paddle.trainer.SGD(cost=cost,
+                                 parameters=parameters,
+                                 update_equation=optimizer,
+                                 extra_layers=extra_layers)
+
+    # End batch and end pass event handler
+    def event_handler(event):
+        if isinstance(event, paddle.event.EndIteration):
+            if event.batch_id % 1 == 0:
+                print "\nPass %d, Batch %d, Cost %f, %s" % (
+                    event.pass_id, event.batch_id, event.cost, event.metrics)
+        if isinstance(event, paddle.event.EndPass):
+            with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
+                trainer.save_parameter_to_tar(f)
+
+            result = trainer.test(reader=test_reader)
+            print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
+
+    trainer.train(
+        reader=train_reader, num_passes=200, event_handler=event_handler)
+
+
+if __name__ == '__main__':
+    main()
\ No newline at end of file

fluid/DeepASR/examples/aishell/profile.sh

+export CUDA_VISIBLE_DEVICES=2,3,4,5
+python -u ../../tools/profile.py --feature_lst data/train_feature.lst \
+                   --label_lst data/train_label.lst \
+                   --mean_var data/aishell/global_mean_var \
+                   --parallel \
+                   --frame_dim 2640  \
+                   --class_num 101  \

fluid/DeepQNetwork/DQN.py

+#-*- coding: utf-8 -*-
+#File: DQN.py
+
+from agent import Model
+import gym
+import argparse
+from tqdm import tqdm
+from expreplay import ReplayMemory, Experience
+import numpy as np
+import os
+
+UPDATE_FREQ = 4
+
+MEMORY_WARMUP_SIZE = 1000
+
+
+def run_episode(agent, env, exp, train_or_test):
+    assert train_or_test in ['train', 'test'], train_or_test
+    total_reward = 0
+    state = env.reset()
+    for step in range(200):
+        action = agent.act(state, train_or_test)
+        next_state, reward, isOver, _ = env.step(action)
+        if train_or_test == 'train':
+            exp.append(Experience(state, action, reward, isOver))
+            # train model
+            # start training 
+            if len(exp) > MEMORY_WARMUP_SIZE:
+                batch_idx = np.random.randint(
+                    len(exp) - 1, size=(args.batch_size))
+                if step % UPDATE_FREQ == 0:
+                    batch_state, batch_action, batch_reward, \
+                    batch_next_state, batch_isOver = exp.sample(batch_idx)
+                    agent.train(batch_state, batch_action, batch_reward, \
+                                batch_next_state, batch_isOver)
+        total_reward += reward
+        state = next_state
+        if isOver:
+            break
+    return total_reward
+
+
+def train_agent():
+    env = gym.make(args.env)
+    state_shape = env.observation_space.shape
+    exp = ReplayMemory(args.mem_size, state_shape)
+    action_dim = env.action_space.n
+    agent = Model(state_shape[0], action_dim, gamma=0.99)
+
+    while len(exp) < MEMORY_WARMUP_SIZE:
+        run_episode(agent, env, exp, train_or_test='train')
+
+    max_episode = 4000
+
+    # train
+    total_episode = 0
+    pbar = tqdm(total=max_episode)
+    recent_100_reward = []
+    for episode in xrange(max_episode):
+        # start epoch
+        total_reward = run_episode(agent, env, exp, train_or_test='train')
+        pbar.set_description('[train]exploration:{}'.format(agent.exploration))
+        pbar.update()
+
+        # recent 100 reward
+        total_reward = run_episode(agent, env, exp, train_or_test='test')
+        recent_100_reward.append(total_reward)
+        if len(recent_100_reward) > 100:
+            recent_100_reward = recent_100_reward[1:]
+        pbar.write("episode:{}    test_reward:{}".format(\
+                    episode, np.mean(recent_100_reward)))
+
+    pbar.close()
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--env', type=str, default='MountainCar-v0', \
+                        help='enviroment to train DQN model, e.g CartPole-v0')
+    parser.add_argument('--gamma', type=float, default=0.99, \
+                        help='discount factor for accumulated reward computation')
+    parser.add_argument('--mem_size', type=int, default=500000, \
+                        help='memory size for experience replay')
+    parser.add_argument('--batch_size', type=int, default=192, \
+                        help='batch size for training')
+    args = parser.parse_args()
+
+    train_agent()

fluid/DeepQNetwork/README.md

+<img src="mountain_car.gif" width="300" height="200">
+
+# Reproduce DQN model
+ + DQN in:
+[Human-level Control Through Deep Reinforcement Learning](http://www.nature.com/nature/journal/v518/n7540/full/nature14236.html)
+
+# Mountain-CAR benchmark & performance
+[MountainCar-v0](https://gym.openai.com/envs/MountainCar-v0/)
+
+A car is on a one-dimensional track, positioned between two "mountains". The goal is to drive up the mountain on the right; however, the car's engine is not strong enough to scale the mountain in a single pass. Therefore, the only way to succeed is to drive back and forth to build up momentum.
+
+
+
+<img src="curve.png" >
+
+
+
+# How to use
++ Dependencies:
+   + python2.7
+   + gym
+   + tqdm
+   + paddle-fluid
++ Start Training:
+   ```
+   # use mountain-car enviroment as default
+   python DQN.py
+
+   # use other enviorment
+   python DQN.py --env CartPole-v0
+   ```

fluid/DeepQNetwork/agent.py

+#-*- coding: utf-8 -*-
+#File: agent.py
+
+import paddle.fluid as fluid
+from paddle.fluid.param_attr import ParamAttr
+import numpy as np
+from tqdm import tqdm
+import math
+
+UPDATE_TARGET_STEPS = 200
+
+
+class Model(object):
+    def __init__(self, state_dim, action_dim, gamma):
+        self.global_step = 0
+        self.state_dim = state_dim
+        self.action_dim = action_dim
+        self.gamma = gamma
+        self.exploration = 1.0
+
+        self._build_net()
+
+    def _get_inputs(self):
+        return [fluid.layers.data(\
+                    name='state', shape=[self.state_dim], dtype='float32'),
+                fluid.layers.data(\
+                    name='action', shape=[1], dtype='int32'),
+                fluid.layers.data(\
+                    name='reward', shape=[], dtype='float32'),
+                fluid.layers.data(\
+                    name='next_s', shape=[self.state_dim], dtype='float32'),
+                fluid.layers.data(\
+                  name='isOver', shape=[], dtype='bool')]
+
+    def _build_net(self):
+        state, action, reward, next_s, isOver = self._get_inputs()
+        self.pred_value = self.get_DQN_prediction(state)
+        self.predict_program = fluid.default_main_program().clone()
+
+        action_onehot = fluid.layers.one_hot(action, self.action_dim)
+        action_onehot = fluid.layers.cast(action_onehot, dtype='float32')
+
+        pred_action_value = fluid.layers.reduce_sum(\
+                    fluid.layers.elementwise_mul(action_onehot, self.pred_value), dim=1)
+
+        targetQ_predict_value = self.get_DQN_prediction(next_s, target=True)
+        best_v = fluid.layers.reduce_max(targetQ_predict_value, dim=1)
+        best_v.stop_gradient = True
+
+        target = reward + (1.0 - fluid.layers.cast(\
+            isOver, dtype='float32')) * self.gamma * best_v
+        cost = fluid.layers.square_error_cost(\
+            input=pred_action_value, label=target)
+        cost = fluid.layers.reduce_mean(cost)
+
+        self._sync_program = self._build_sync_target_network()
+
+        optimizer = fluid.optimizer.Adam(1e-3)
+        optimizer.minimize(cost)
+
+        # define program
+        self.train_program = fluid.default_main_program()
+
+        # fluid exe
+        place = fluid.CUDAPlace(0)
+        self.exe = fluid.Executor(place)
+        self.exe.run(fluid.default_startup_program())
+
+    def get_DQN_prediction(self, state, target=False):
+        variable_field = 'target' if target else 'policy'
+        # layer fc1
+        param_attr = ParamAttr(name='{}_fc1'.format(variable_field))
+        bias_attr = ParamAttr(name='{}_fc1_b'.format(variable_field))
+        fc1 = fluid.layers.fc(input=state,
+                              size=256,
+                              act='relu',
+                              param_attr=param_attr,
+                              bias_attr=bias_attr)
+
+        param_attr = ParamAttr(name='{}_fc2'.format(variable_field))
+        bias_attr = ParamAttr(name='{}_fc2_b'.format(variable_field))
+        fc2 = fluid.layers.fc(input=fc1,
+                              size=128,
+                              act='tanh',
+                              param_attr=param_attr,
+                              bias_attr=bias_attr)
+
+        param_attr = ParamAttr(name='{}_fc3'.format(variable_field))
+        bias_attr = ParamAttr(name='{}_fc3_b'.format(variable_field))
+        value = fluid.layers.fc(input=fc2,
+                                size=self.action_dim,
+                                param_attr=param_attr,
+                                bias_attr=bias_attr)
+
+        return value
+
+    def _build_sync_target_network(self):
+        vars = fluid.default_main_program().list_vars()
+        policy_vars = []
+        target_vars = []
+        for var in vars:
+            if 'GRAD' in var.name: continue
+            if 'policy' in var.name:
+                policy_vars.append(var)
+            elif 'target' in var.name:
+                target_vars.append(var)
+
+        policy_vars.sort(key=lambda x: x.name.split('policy_')[1])
+        target_vars.sort(key=lambda x: x.name.split('target_')[1])
+
+        sync_program = fluid.default_main_program().clone()
+        with fluid.program_guard(sync_program):
+            sync_ops = []
+            for i, var in enumerate(policy_vars):
+                sync_op = fluid.layers.assign(policy_vars[i], target_vars[i])
+                sync_ops.append(sync_op)
+        sync_program = sync_program.prune(sync_ops)
+        return sync_program
+
+    def act(self, state, train_or_test):
+        sample = np.random.random()
+        if train_or_test == 'train' and sample < self.exploration:
+            act = np.random.randint(self.action_dim)
+        else:
+            state = np.expand_dims(state, axis=0)
+            pred_Q = self.exe.run(self.predict_program,
+                                  feed={'state': state.astype('float32')},
+                                  fetch_list=[self.pred_value])[0]
+            pred_Q = np.squeeze(pred_Q, axis=0)
+            act = np.argmax(pred_Q)
+        self.exploration = max(0.1, self.exploration - 1e-6)
+        return act
+
+    def train(self, state, action, reward, next_state, isOver):
+        if self.global_step % UPDATE_TARGET_STEPS == 0:
+            self.sync_target_network()
+        self.global_step += 1
+
+        action = np.expand_dims(action, -1)
+        self.exe.run(self.train_program, \
+                  feed={'state': state, \
+                        'action': action, \
+                        'reward': reward, \
+                        'next_s': next_state, \
+                        'isOver': isOver})
+
+    def sync_target_network(self):
+        self.exe.run(self._sync_program)

fluid/DeepQNetwork/expreplay.py

+#-*- coding: utf-8 -*-
+#File: expreplay.py
+
+from collections import namedtuple
+import numpy as np
+
+Experience = namedtuple('Experience', ['state', 'action', 'reward', 'isOver'])
+
+
+class ReplayMemory(object):
+    def __init__(self, max_size, state_shape):
+        self.max_size = int(max_size)
+        self.state_shape = state_shape
+
+        self.state = np.zeros((self.max_size, ) + state_shape, dtype='float32')
+        self.action = np.zeros((self.max_size, ), dtype='int32')
+        self.reward = np.zeros((self.max_size, ), dtype='float32')
+        self.isOver = np.zeros((self.max_size, ), dtype='bool')
+
+        self._curr_size = 0
+        self._curr_pos = 0
+
+    def append(self, exp):
+        if self._curr_size < self.max_size:
+            self._assign(self._curr_pos, exp)
+            self._curr_size += 1
+        else:
+            self._assign(self._curr_pos, exp)
+        self._curr_pos = (self._curr_pos + 1) % self.max_size
+
+    def _assign(self, pos, exp):
+        self.state[pos] = exp.state
+        self.action[pos] = exp.action
+        self.reward[pos] = exp.reward
+        self.isOver[pos] = exp.isOver
+
+    def __len__(self):
+        return self._curr_size
+
+    def sample(self, batch_idx):
+        # index mapping to avoid sampling lastest state
+        batch_idx = (self._curr_pos + batch_idx) % self._curr_size
+        next_idx = (batch_idx + 1) % self._curr_size
+
+        state = self.state[batch_idx]
+        reward = self.reward[batch_idx]
+        action = self.action[batch_idx]
+        next_state = self.state[next_idx]
+        isOver = self.isOver[batch_idx]
+        return (state, action, reward, next_state, isOver)

fluid/ocr_recognition/ctc_reader.py

@@ -136,6 +136,7 @@ class DataGenerator(object):
                     img = Image.open(img_path).convert('L')
                     img = np.array(img) - 127.5
                     img = img[np.newaxis, ...]
+                    label = [int(c) for c in line.split(' ')[3].split(',')]
                     yield img, label
             else:
                 while True:

generate_sequence_by_rnn_lm/README.md

@@ -99,7 +99,7 @@ RNN是一个序列模型，基本思路是：在时刻$t$，将前一时刻$t-1$
     ```
     1. `max_word_num`：指定字典中含有多少个词。
     2. `cutoff_word_fre`：字典中词语在训练语料中出现的最低频率。
-- 加入指定了 `max_word_num = 5000`，并且 `cutoff_word_fre = 10`，词频统计发现训练语料中出现频率高于10次的词语仅有3000个，那么最终会取3000个词构成词典。
+- 假如指定了 `max_word_num = 5000`，并且 `cutoff_word_fre = 10`，词频统计发现训练语料中出现频率高于10次的词语仅有3000个，那么最终会取3000个词构成词典。
 - 构建词典时，会自动加入两个特殊符号：
     1. `<unk>`：不出现在字典中的词
     2. `<e>`：句子的结束符