diff --git a/01.fit_a_line/README.cn.md b/01.fit_a_line/README.cn.md index 936142bfbc427fa085797f2e62b136709e085405..a969c752767a575d78e86bd51d763702b91eb551 100644 --- a/01.fit_a_line/README.cn.md +++ b/01.fit_a_line/README.cn.md @@ -100,6 +100,7 @@ $$MSE=\frac{1}{n}\sum_{i=1}^{n}{(\hat{Y_i}-Y_i)}^2$$ import paddle import paddle.fluid as fluid import numpy +from __future__ import print_function ``` 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) @@ -179,7 +180,7 @@ feed_order=['x', 'y'] 除此之外，可以定义一个事件相应器来处理类似`打印训练进程`的事件： ```python -# Specify the directory path to save the parameters +# Specify the directory to save the parameters params_dirname = "fit_a_line.inference.model" # Plot data @@ -190,11 +191,11 @@ plot_cost = Ploter(train_title, test_title) step = 0 -# event_handler to print training and testing info +# event_handler prints training and testing info def event_handler_plot(event): global step if isinstance(event, fluid.EndStepEvent): - if event.step % 10 == 0: # every 10 batches, record a test cost + if event.step % 10 == 0: # record the test cost every 10 seconds test_metrics = trainer.test( reader=test_reader, feed_order=feed_order) @@ -251,10 +252,20 @@ inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) batch_size = 10 -tensor_x = numpy.random.uniform(0, 10, [batch_size, 13]).astype("float32") +test_reader = paddle.batch(paddle.dataset.uci_housing.test(),batch_size=batch_size) +test_data = test_reader().next() +test_feat = numpy.array([data[0] for data in test_data]).astype("float32") +test_label = numpy.array([data[1] for data in test_data]).astype("float32") -results = inferencer.infer({'x': tensor_x}) -print("infer results: ", results[0]) +results = inferencer.infer({'x': test_feat}) + +print("infer results: (House Price)") +for k in range(0, batch_size-1): + print("%d. %f" % (k, results[0][k])) + +print("\nground truth:") +for k in range(0, batch_size-1): + print("%d. %f" % (k, test_label[k])) ``` ## 总结 diff --git a/01.fit_a_line/README.md b/01.fit_a_line/README.md index e672caf1206b13b50abbd57eb2fc465fb656cce1..407d0411eb7a2f3cd990618a22b34eaca0abfd28 100644 --- a/01.fit_a_line/README.md +++ b/01.fit_a_line/README.md @@ -108,6 +108,7 @@ Our program starts with importing necessary packages: import paddle import paddle.fluid as fluid import numpy +from __future__ import print_function ``` We encapsulated the [UCI Housing Data Set](https://archive.ics.uci.edu/ml/datasets/Housing) in our Python module `uci_housing`. This module can @@ -189,7 +190,7 @@ feed_order=['x', 'y'] Moreover, an event handler is provided to print the training progress: ```python -# Specify the directory path to save the parameters +# Specify the directory to save the parameters params_dirname = "fit_a_line.inference.model" # Plot data @@ -200,11 +201,11 @@ plot_cost = Ploter(train_title, test_title) step = 0 -# event_handler to print training and testing info +# event_handler prints training and testing info def event_handler_plot(event): global step if isinstance(event, fluid.EndStepEvent): - if event.step % 10 == 0: # every 10 batches, record a test cost + if event.step % 10 == 0: #record a test cost every 10 batches test_metrics = trainer.test( reader=test_reader, feed_order=feed_order) @@ -263,10 +264,20 @@ inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) batch_size = 10 -tensor_x = numpy.random.uniform(0, 10, [batch_size, 13]).astype("float32") +test_reader = paddle.batch(paddle.dataset.uci_housing.test(),batch_size=batch_size) +test_data = test_reader().next() +test_feat = numpy.array([data[0] for data in test_data]).astype("float32") +test_label = numpy.array([data[1] for data in test_data]).astype("float32") -results = inferencer.infer({'x': tensor_x}) -print("infer results: ", results[0]) +results = inferencer.infer({'x': test_feat}) + +print("infer results: (House Price)") +for k in range(0, batch_size-1): + print("%d. %f" % (k, results[0][k])) + +print("\nground truth:") +for k in range(0, batch_size-1): + print("%d. %f" % (k, test_label[k])) ``` ## Summary diff --git a/01.fit_a_line/image/ranges.png b/01.fit_a_line/image/ranges.png index 5d86b12715f46afbafb7d50e2938e184219b5b95..5325df4800985983e17476f007658d1cdb170b1c 100644 Binary files a/01.fit_a_line/image/ranges.png and b/01.fit_a_line/image/ranges.png differ diff --git a/01.fit_a_line/index.cn.html b/01.fit_a_line/index.cn.html index 32760a22092f136746a72328b48502548caf7692..9a0b9918dcbbb31c4615908b3cc323eb59b36835 100644 --- a/01.fit_a_line/index.cn.html +++ b/01.fit_a_line/index.cn.html @@ -142,6 +142,7 @@ $$MSE=\frac{1}{n}\sum_{i=1}^{n}{(\hat{Y_i}-Y_i)}^2$$ import paddle import paddle.fluid as fluid import numpy +from __future__ import print_function ``` 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) @@ -221,7 +222,7 @@ feed_order=['x', 'y'] 除此之外，可以定义一个事件相应器来处理类似`打印训练进程`的事件： ```python -# Specify the directory path to save the parameters +# Specify the directory to save the parameters params_dirname = "fit_a_line.inference.model" # Plot data @@ -232,11 +233,11 @@ plot_cost = Ploter(train_title, test_title) step = 0 -# event_handler to print training and testing info +# event_handler prints training and testing info def event_handler_plot(event): global step if isinstance(event, fluid.EndStepEvent): - if event.step % 10 == 0: # every 10 batches, record a test cost + if event.step % 10 == 0: # record the test cost every 10 seconds test_metrics = trainer.test( reader=test_reader, feed_order=feed_order) @@ -293,10 +294,20 @@ inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) batch_size = 10 -tensor_x = numpy.random.uniform(0, 10, [batch_size, 13]).astype("float32") +test_reader = paddle.batch(paddle.dataset.uci_housing.test(),batch_size=batch_size) +test_data = test_reader().next() +test_feat = numpy.array([data[0] for data in test_data]).astype("float32") +test_label = numpy.array([data[1] for data in test_data]).astype("float32") -results = inferencer.infer({'x': tensor_x}) -print("infer results: ", results[0]) +results = inferencer.infer({'x': test_feat}) + +print("infer results: (House Price)") +for k in range(0, batch_size-1): + print("%d. %f" % (k, results[0][k])) + +print("\nground truth:") +for k in range(0, batch_size-1): + print("%d. %f" % (k, test_label[k])) ``` ## 总结 diff --git a/01.fit_a_line/index.html b/01.fit_a_line/index.html index cab234ff49188aa0be1941c5c81c4c1fc3a9d2a3..4c6cf7f431c599abdae0f6e4109ca571ba613f4d 100644 --- a/01.fit_a_line/index.html +++ b/01.fit_a_line/index.html @@ -150,6 +150,7 @@ Our program starts with importing necessary packages: import paddle import paddle.fluid as fluid import numpy +from __future__ import print_function ``` We encapsulated the [UCI Housing Data Set](https://archive.ics.uci.edu/ml/datasets/Housing) in our Python module `uci_housing`. This module can @@ -231,7 +232,7 @@ feed_order=['x', 'y'] Moreover, an event handler is provided to print the training progress: ```python -# Specify the directory path to save the parameters +# Specify the directory to save the parameters params_dirname = "fit_a_line.inference.model" # Plot data @@ -242,11 +243,11 @@ plot_cost = Ploter(train_title, test_title) step = 0 -# event_handler to print training and testing info +# event_handler prints training and testing info def event_handler_plot(event): global step if isinstance(event, fluid.EndStepEvent): - if event.step % 10 == 0: # every 10 batches, record a test cost + if event.step % 10 == 0: #record a test cost every 10 batches test_metrics = trainer.test( reader=test_reader, feed_order=feed_order) @@ -305,10 +306,20 @@ inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) batch_size = 10 -tensor_x = numpy.random.uniform(0, 10, [batch_size, 13]).astype("float32") +test_reader = paddle.batch(paddle.dataset.uci_housing.test(),batch_size=batch_size) +test_data = test_reader().next() +test_feat = numpy.array([data[0] for data in test_data]).astype("float32") +test_label = numpy.array([data[1] for data in test_data]).astype("float32") -results = inferencer.infer({'x': tensor_x}) -print("infer results: ", results[0]) +results = inferencer.infer({'x': test_feat}) + +print("infer results: (House Price)") +for k in range(0, batch_size-1): + print("%d. %f" % (k, results[0][k])) + +print("\nground truth:") +for k in range(0, batch_size-1): + print("%d. %f" % (k, test_label[k])) ``` ## Summary diff --git a/03.image_classification/README.cn.md b/03.image_classification/README.cn.md index 59180d1db79c88e1986606cf06873559a2b7b961..3f21327ba257806aef5f7464d0708a95421b0d4a 100644 --- a/03.image_classification/README.cn.md +++ b/03.image_classification/README.cn.md @@ -160,6 +160,7 @@ import paddle import paddle.fluid as fluid import numpy import sys +from __future__ import print_function ``` 本教程中我们提供了VGG和ResNet两个模型的配置。 @@ -426,7 +427,7 @@ def event_handler(event): 通过`trainer.train`函数训练: -**注意:** CPU，每个 Epoch 将花费大约15～20分钟。这部分可能需要一段时间。请随意修改代码，在GPU上运行测试，以提高培训速度。 +**注意:** CPU，每个 Epoch 将花费大约15～20分钟。这部分可能需要一段时间。请随意修改代码，在GPU上运行测试，以提高训练速度。 ```python trainer.train( @@ -499,10 +500,10 @@ img = load_image(cur_dir + '/image/dog.png') ```python inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) - +label_list = ["airplane", "automobile", "bird", "cat", "deer", "dog", "frog", "horse", "ship", "truck"] # inference results = inferencer.infer({'pixel': img}) -print("infer results: ", results) +print("infer results: %s" % label_list[np.argmax(results[0])]) ``` ## 总结 diff --git a/03.image_classification/README.md b/03.image_classification/README.md index f24fbdeb9458f1730fe64bd6d54b2c71400d4a4e..c3632e9e9a85e66ea8b7ab4ed14ed635c72c9258 100644 --- a/03.image_classification/README.md +++ b/03.image_classification/README.md @@ -171,6 +171,7 @@ import paddle import paddle.fluid as fluid import numpy import sys +from __future__ import print_function ``` Now we are going to walk you through the implementations of the VGG and ResNet. @@ -514,9 +515,10 @@ Now we are ready to do inference. inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) +label_list = ["airplane", "automobile", "bird", "cat", "deer", "dog", "frog", "horse", "ship", "truck"] # inference results = inferencer.infer({'pixel': img}) -print("infer results: ", results) +print("infer results: %s" % label_list[np.argmax(results[0])]) ``` diff --git a/03.image_classification/index.cn.html b/03.image_classification/index.cn.html index dbd1f1c5978e91126104309e67527b9243a0fc75..0aa3cf2554cbeb07b57388e405fc0da0bba50665 100644 --- a/03.image_classification/index.cn.html +++ b/03.image_classification/index.cn.html @@ -202,6 +202,7 @@ import paddle import paddle.fluid as fluid import numpy import sys +from __future__ import print_function ``` 本教程中我们提供了VGG和ResNet两个模型的配置。 @@ -468,7 +469,7 @@ def event_handler(event): 通过`trainer.train`函数训练: -**注意:** CPU，每个 Epoch 将花费大约15～20分钟。这部分可能需要一段时间。请随意修改代码，在GPU上运行测试，以提高培训速度。 +**注意:** CPU，每个 Epoch 将花费大约15～20分钟。这部分可能需要一段时间。请随意修改代码，在GPU上运行测试，以提高训练速度。 ```python trainer.train( @@ -541,10 +542,10 @@ img = load_image(cur_dir + '/image/dog.png') ```python inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) - +label_list = ["airplane", "automobile", "bird", "cat", "deer", "dog", "frog", "horse", "ship", "truck"] # inference results = inferencer.infer({'pixel': img}) -print("infer results: ", results) +print("infer results: %s" % label_list[np.argmax(results[0])]) ``` ## 总结 diff --git a/03.image_classification/index.html b/03.image_classification/index.html index 0a37e7016c2849bfe9b1c6bf4c1acda864a0b1d0..6c2f1f6643caaab13831f824fbbfdb85644c76ec 100644 --- a/03.image_classification/index.html +++ b/03.image_classification/index.html @@ -213,6 +213,7 @@ import paddle import paddle.fluid as fluid import numpy import sys +from __future__ import print_function ``` Now we are going to walk you through the implementations of the VGG and ResNet. @@ -556,9 +557,10 @@ Now we are ready to do inference. inferencer = fluid.Inferencer( infer_func=inference_program, param_path=params_dirname, place=place) +label_list = ["airplane", "automobile", "bird", "cat", "deer", "dog", "frog", "horse", "ship", "truck"] # inference results = inferencer.infer({'pixel': img}) -print("infer results: ", results) +print("infer results: %s" % label_list[np.argmax(results[0])]) ``` diff --git a/04.word2vec/README.cn.md b/04.word2vec/README.cn.md index 00493fdf163bfbb8610b4378e72484ed3e3b9a71..450d469334f62120b4237699c934d5d8565059d9 100644 --- a/04.word2vec/README.cn.md +++ b/04.word2vec/README.cn.md @@ -204,6 +204,7 @@ from functools import partial import math import os import sys +from __future__ import print_function ``` 然后，定义参数： @@ -311,7 +312,7 @@ def train(use_cuda, train_program, params_dirname): feed_order=['firstw', 'secondw', 'thirdw', 'fourthw', 'nextw']) avg_cost = outs[0] - print "Step %d: Average Cost %f" % (event.step, avg_cost) + print("Step %d: Average Cost %f" % (event.step, avg_cost)) # If average cost is lower than 5.8, we consider the model good enough to stop. # Note 5.8 is a relatively high value. In order to get a better model, one should @@ -337,7 +338,7 @@ def train(use_cuda, train_program, params_dirname): - `trainer.train`将会开始训练。从`event_handler`返回的监控情况如下： -```python +```text Step 0: Average Cost 7.337213 Step 10: Average Cost 6.136128 Step 20: Average Cost 5.766995 @@ -394,8 +395,7 @@ def infer(use_cuda, inference_program, params_dirname=None): 在经历3分钟的短暂训练后，我们得到如下的预测。我们的模型预测 `among a group of` 的下一个词是`a`。这比较符合文法规律。如果我们训练时间更长，比如几个小时，那么我们会得到的下一个预测是 `workers`。 - -```python +```text [[0.00106646 0.0007907 0.00072041 ... 0.00049024 0.00041355 0.00084464]] 6 a diff --git a/04.word2vec/README.md b/04.word2vec/README.md index e99d5336b64dd57dec34e704256aef4eb36c7db9..330159055b681368b3d5271ecb8a4f36c5ea0eb6 100644 --- a/04.word2vec/README.md +++ b/04.word2vec/README.md @@ -222,6 +222,7 @@ from functools import partial import math import os import sys +from __future__ import print_function ``` - Configure parameters and build word dictionary. @@ -328,7 +329,7 @@ def train(use_cuda, train_program, params_dirname): # We output cost every 10 steps. if event.step % 10 == 0: - print "Step %d: Average Cost %f" % (event.step, avg_cost) + print("Step %d: Average Cost %f" % (event.step, avg_cost)) # If average cost is lower than 5.8, we consider the model good enough to stop. # Note 5.8 is a relatively high value. In order to get a better model, one should diff --git a/04.word2vec/index.cn.html b/04.word2vec/index.cn.html index e146f6a0f47292134930f4577a57b0fbd9c9ab5b..6f0ec9ce65250c9b3b162375eeb4be9e47714080 100644 --- a/04.word2vec/index.cn.html +++ b/04.word2vec/index.cn.html @@ -246,6 +246,7 @@ from functools import partial import math import os import sys +from __future__ import print_function ``` 然后，定义参数： @@ -353,7 +354,7 @@ def train(use_cuda, train_program, params_dirname): feed_order=['firstw', 'secondw', 'thirdw', 'fourthw', 'nextw']) avg_cost = outs[0] - print "Step %d: Average Cost %f" % (event.step, avg_cost) + print("Step %d: Average Cost %f" % (event.step, avg_cost)) # If average cost is lower than 5.8, we consider the model good enough to stop. # Note 5.8 is a relatively high value. In order to get a better model, one should @@ -379,7 +380,7 @@ def train(use_cuda, train_program, params_dirname): - `trainer.train`将会开始训练。从`event_handler`返回的监控情况如下： -```python +```text Step 0: Average Cost 7.337213 Step 10: Average Cost 6.136128 Step 20: Average Cost 5.766995 @@ -436,8 +437,7 @@ def infer(use_cuda, inference_program, params_dirname=None): 在经历3分钟的短暂训练后，我们得到如下的预测。我们的模型预测 `among a group of` 的下一个词是`a`。这比较符合文法规律。如果我们训练时间更长，比如几个小时，那么我们会得到的下一个预测是 `workers`。 - -```python +```text [[0.00106646 0.0007907 0.00072041 ... 0.00049024 0.00041355 0.00084464]] 6 a diff --git a/04.word2vec/index.html b/04.word2vec/index.html index 4208db87c3a9820ad5ab6f912cd27184c61b8d34..8b324dab8e18745289f4a1ec8db7d505c821afac 100644 --- a/04.word2vec/index.html +++ b/04.word2vec/index.html @@ -264,6 +264,7 @@ from functools import partial import math import os import sys +from __future__ import print_function ``` - Configure parameters and build word dictionary. @@ -370,7 +371,7 @@ def train(use_cuda, train_program, params_dirname): # We output cost every 10 steps. if event.step % 10 == 0: - print "Step %d: Average Cost %f" % (event.step, avg_cost) + print("Step %d: Average Cost %f" % (event.step, avg_cost)) # If average cost is lower than 5.8, we consider the model good enough to stop. # Note 5.8 is a relatively high value. In order to get a better model, one should diff --git a/05.recommender_system/README.cn.md b/05.recommender_system/README.cn.md index 7b4e1092c492910bcb85aa6e934c3f65b1cad819..160e24525d274df0dd7c6b794e6714dfa400ac79 100644 --- a/05.recommender_system/README.cn.md +++ b/05.recommender_system/README.cn.md @@ -215,6 +215,7 @@ print "User %s rates Movie %s with Score %s"%(user_info[uid], movie_info[mov_id] ```python +from __future__ import print_function import math import sys import numpy as np @@ -508,6 +509,11 @@ results = inferencer.infer( 'movie_title': movie_title }, return_numpy=False) + +predict_rating = np.array(results[0]) +print("Predict Rating of user id 1 on movie \"" + infer_movie_name + "\" is " + str(predict_rating[0][0])) +print("Actual Rating of user id 1 on movie \"" + infer_movie_name + "\" is 4.") + ``` ## 总结 diff --git a/05.recommender_system/README.md b/05.recommender_system/README.md index 2b1ed98e6747980921d63e5a33b6584063adb13b..939c5a433afa7fc358323f1daf8312f90b292137 100644 --- a/05.recommender_system/README.md +++ b/05.recommender_system/README.md @@ -185,6 +185,7 @@ After issuing a command `python train.py`, training will start immediately. The Our program starts with importing necessary packages and initializing some global variables: ```python +from __future__ import print_function import math import sys import numpy as np diff --git a/05.recommender_system/index.cn.html b/05.recommender_system/index.cn.html index 2fc69601d9a70e6e9f81231311e669ead5d2fc17..1180a7e512f836ef6ac0c4a8b52141df79da78c4 100644 --- a/05.recommender_system/index.cn.html +++ b/05.recommender_system/index.cn.html @@ -257,6 +257,7 @@ print "User %s rates Movie %s with Score %s"%(user_info[uid], movie_info[mov_id] ```python +from __future__ import print_function import math import sys import numpy as np @@ -550,6 +551,11 @@ results = inferencer.infer( 'movie_title': movie_title }, return_numpy=False) + +predict_rating = np.array(results[0]) +print("Predict Rating of user id 1 on movie \"" + infer_movie_name + "\" is " + str(predict_rating[0][0])) +print("Actual Rating of user id 1 on movie \"" + infer_movie_name + "\" is 4.") + ``` ## 总结 diff --git a/05.recommender_system/index.html b/05.recommender_system/index.html index 9aca3c3894cf6ed60f9386c022abb88022fcde86..461923573e3f04a38e31de86e1da858bde30466e 100644 --- a/05.recommender_system/index.html +++ b/05.recommender_system/index.html @@ -227,6 +227,7 @@ After issuing a command `python train.py`, training will start immediately. The Our program starts with importing necessary packages and initializing some global variables: ```python +from __future__ import print_function import math import sys import numpy as np diff --git a/05.recommender_system/train.py b/05.recommender_system/train.py index b04a28255a77013f8b55eccbb84f8b0a5ad8e295..564abd33d89da61cd65fee3fda2684d9ac152f90 100644 --- a/05.recommender_system/train.py +++ b/05.recommender_system/train.py @@ -12,6 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. +from __future__ import print_function import math import sys import numpy as np @@ -233,7 +234,11 @@ def infer(use_cuda, inference_program, params_dirname): }, return_numpy=False) - print("infer results: ", np.array(results[0])) + predict_rating = np.array(results[0]) + print("Predict Rating of user id 1 on movie \"" + infer_movie_name + + "\" is " + str(predict_rating[0][0])) + print("Actual Rating of user id 1 on movie \"" + infer_movie_name + + "\" is 4.") def main(use_cuda): diff --git a/06.understand_sentiment/README.cn.md b/06.understand_sentiment/README.cn.md index 31c80285f1dcf36fca83a1130dd268973781dacb..88097dd3fbf3c79694136ffa8d80bfbaa3cc7da0 100644 --- a/06.understand_sentiment/README.cn.md +++ b/06.understand_sentiment/README.cn.md @@ -107,6 +107,7 @@ Paddle在`dataset/imdb.py`中提实现了imdb数据集的自动下载和读取 在该示例中，我们实现了两种文本分类算法，分别基于[推荐系统](https://github.com/PaddlePaddle/book/tree/develop/05.recommender_system)一节介绍过的文本卷积神经网络，以及[栈式双向LSTM](#栈式双向LSTM（Stacked Bidirectional LSTM）)。我们首先引入要用到的库和定义全局变量： ```python +from __future__ import print_function import paddle import paddle.fluid as fluid from functools import partial @@ -115,6 +116,7 @@ import numpy as np CLASS_DIM = 2 EMB_DIM = 128 HID_DIM = 512 +STACKED_NUM = 3 BATCH_SIZE = 128 USE_GPU = False ``` @@ -168,17 +170,12 @@ def stacked_lstm_net(data, input_dim, class_dim, emb_dim, hid_dim, stacked_num): input=fc, size=hid_dim, is_reverse=(i % 2) == 0) 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 + 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 ``` 以上的栈式双向LSTM抽象出了高级特征并把其映射到和分类类别数同样大小的向量上。`paddle.activation.Softmax`函数用来计算分类属于某个类别的概率。 @@ -193,6 +190,7 @@ def inference_program(word_dict): dict_dim = len(word_dict) net = convolution_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM) + # net = stacked_lstm_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM, STACKED_NUM) return net ``` @@ -301,7 +299,7 @@ trainer.train( ```python inferencer = fluid.Inferencer( - inference_program, param_path=params_dirname, place=place) + infer_func=partial(inference_program, word_dict), param_path=params_dirname, place=place) ``` ### 生成测试用输入数据 diff --git a/06.understand_sentiment/README.md b/06.understand_sentiment/README.md index f427f134de8ee0e3c9844869bb1961b6235b6c20..959d7ad600969bfc4f372cd3b582e2482cf6039e 100644 --- a/06.understand_sentiment/README.md +++ b/06.understand_sentiment/README.md @@ -103,6 +103,7 @@ After issuing a command `python train.py`, training will start immediately. The Our program starts with importing necessary packages and initializing some global variables: ```python +from __future__ import print_function import paddle import paddle.fluid as fluid from functools import partial @@ -111,6 +112,7 @@ import numpy as np CLASS_DIM = 2 EMB_DIM = 128 HID_DIM = 512 +STACKED_NUM = 3 BATCH_SIZE = 128 USE_GPU = False ``` @@ -192,6 +194,7 @@ def inference_program(word_dict): dict_dim = len(word_dict) net = convolution_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM) + # net = stacked_lstm_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM, STACKED_NUM) return net ``` diff --git a/06.understand_sentiment/index.cn.html b/06.understand_sentiment/index.cn.html index 2be442eabded20f444ef8962eb25b49aab7c213c..f9935e890397460f712769a3fa61314f384a24c1 100644 --- a/06.understand_sentiment/index.cn.html +++ b/06.understand_sentiment/index.cn.html @@ -149,6 +149,7 @@ Paddle在`dataset/imdb.py`中提实现了imdb数据集的自动下载和读取 在该示例中，我们实现了两种文本分类算法，分别基于[推荐系统](https://github.com/PaddlePaddle/book/tree/develop/05.recommender_system)一节介绍过的文本卷积神经网络，以及[栈式双向LSTM](#栈式双向LSTM（Stacked Bidirectional LSTM）)。我们首先引入要用到的库和定义全局变量： ```python +from __future__ import print_function import paddle import paddle.fluid as fluid from functools import partial @@ -157,6 +158,7 @@ import numpy as np CLASS_DIM = 2 EMB_DIM = 128 HID_DIM = 512 +STACKED_NUM = 3 BATCH_SIZE = 128 USE_GPU = False ``` @@ -210,17 +212,12 @@ def stacked_lstm_net(data, input_dim, class_dim, emb_dim, hid_dim, stacked_num): input=fc, size=hid_dim, is_reverse=(i % 2) == 0) 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 + 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 ``` 以上的栈式双向LSTM抽象出了高级特征并把其映射到和分类类别数同样大小的向量上。`paddle.activation.Softmax`函数用来计算分类属于某个类别的概率。 @@ -235,6 +232,7 @@ def inference_program(word_dict): dict_dim = len(word_dict) net = convolution_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM) + # net = stacked_lstm_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM, STACKED_NUM) return net ``` @@ -343,7 +341,7 @@ trainer.train( ```python inferencer = fluid.Inferencer( - inference_program, param_path=params_dirname, place=place) + infer_func=partial(inference_program, word_dict), param_path=params_dirname, place=place) ``` ### 生成测试用输入数据 diff --git a/06.understand_sentiment/index.html b/06.understand_sentiment/index.html index 54f0a5b2dd4cd403074f92a967932d081396c303..2afdcfb9cc24a163eb816d4decc605cafb54e12f 100644 --- a/06.understand_sentiment/index.html +++ b/06.understand_sentiment/index.html @@ -145,6 +145,7 @@ After issuing a command `python train.py`, training will start immediately. The Our program starts with importing necessary packages and initializing some global variables: ```python +from __future__ import print_function import paddle import paddle.fluid as fluid from functools import partial @@ -153,6 +154,7 @@ import numpy as np CLASS_DIM = 2 EMB_DIM = 128 HID_DIM = 512 +STACKED_NUM = 3 BATCH_SIZE = 128 USE_GPU = False ``` @@ -234,6 +236,7 @@ def inference_program(word_dict): dict_dim = len(word_dict) net = convolution_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM) + # net = stacked_lstm_net(data, dict_dim, CLASS_DIM, EMB_DIM, HID_DIM, STACKED_NUM) return net ``` diff --git a/07.label_semantic_roles/README.cn.md b/07.label_semantic_roles/README.cn.md index e8b54ceca16fa5d0a1e10f1b1b8eb0fbe1fc22c9..bbfa6cd3e4aca0a21046791000f2c36865a13d9b 100644 --- a/07.label_semantic_roles/README.cn.md +++ b/07.label_semantic_roles/README.cn.md @@ -25,7 +25,7 @@ $$\mbox{[小明]}_{\mbox{Agent}}\mbox{[昨天]}_{\mbox{Time}}\mbox{[晚上]}_\mb 图1. 依存句法分析句法树示例 -然而，完全句法分析需要确定句子所包含的全部句法信息，并确定句子各成分之间的关系，是一个非常困难的任务，目前技术下的句法分析准确率并不高，句法分析的细微错误都会导致SRL的错误。为了降低问题的复杂度，同时获得一定的句法结构信息，“浅层句法分析”的思想应运而生。浅层句法分析也称为部分句法分析（partial parsing）或语块划分（chunking）。和完全句法分析得到一颗完整的句法树不同，浅层句法分析只需要识别句子中某些结构相对简单的独立成分，例如：动词短语，这些被识别出来的结构称为语块。为了回避 “无法获得准确率较高的句法树” 所带来的困难，一些研究\[[1](#参考文献)\]也提出了基于语块（chunk）的SRL方法。基于语块的SRL方法将SRL作为一个序列标注问题来解决。序列标注任务一般都会采用BIO表示方式来定义序列标注的标签集，我们先来介绍这种表示方法。在BIO表示法中，B代表语块的开始，I代表语块的中间，O代表语块结束。通过B、I、O 三种标记将不同的语块赋予不同的标签，例如：对于一个角色为A的论元，将它所包含的第一个语块赋予标签B-A，将它所包含的其它语块赋予标签I-A，不属于任何论元的语块赋予标签O。 +然而，完全句法分析需要确定句子所包含的全部句法信息，并确定句子各成分之间的关系，是一个非常困难的任务，目前技术下的句法分析准确率并不高，句法分析的细微错误都会导致SRL的错误。为了降低问题的复杂度，同时获得一定的句法结构信息，“浅层句法分析”的思想应运而生。浅层句法分析也称为部分句法分析（partial parsing）或语块划分（chunking）。和完全句法分析得到一颗完整的句法树不同，浅层句法分析只需要识别句子中某些结构相对简单的独立成分，例如：动词短语，这些被识别出来的结构称为语块。为了回避 “无法获得准确率较高的句法树” 所带来的困难，一些研究\[[1](#参考文献)\]也提出了基于语块（chunk）的SRL方法。基于语块的SRL方法将SRL作为一个序列标注问题来解决。序列标注任务一般都会采用BIO表示方式来定义序列标注的标签集，我们先来介绍这种表示方法。在BIO表示法中，B代表语块的开始，I代表语块的中间，O代表语块结束。通过B、I、O 三种标记将不同的语块赋予不同的标签，例如：对于一个由角色A拓展得到的语块组，将它所包含的第一个语块赋予标签B-A，将它所包含的其它语块赋予标签I-A，不属于任何论元的语块赋予标签O。 我们继续以上面的这句话为例，图1展示了BIO表示方法。 @@ -151,14 +151,6 @@ conll05st-release/ 4. 构造以BIO法表示的标记； 5. 依据词典获取词对应的整数索引。 - -```python -# import paddle.v2.dataset.conll05 as conll05 -# conll05.corpus_reader函数完成上面第1步和第2步. -# conll05.reader_creator函数完成上面第3步到第5步. -# conll05.test函数可以获取处理之后的每条样本来供PaddlePaddle训练. -``` - 预处理完成之后一条训练样本包含9个特征，分别是：句子序列、谓词、谓词上下文（占 5 列）、谓词上下区域标志、标注序列。下表是一条训练样本的示例。 | 句子序列 | 谓词 | 谓词上下文（窗口 = 5） | 谓词上下文区域标记 | 标注序列 | @@ -187,6 +179,8 @@ conll05st-release/ 获取词典，打印词典大小： ```python +from __future__ import print_function + import math, os import numpy as np import paddle @@ -201,9 +195,9 @@ word_dict_len = len(word_dict) label_dict_len = len(label_dict) pred_dict_len = len(verb_dict) -print word_dict_len -print label_dict_len -print pred_dict_len +print('word_dict_len: ', word_dict_len) +print('label_dict_len: ', label_dict_len) +print('pred_dict_len: ', pred_dict_len) ``` ## 模型配置说明 @@ -431,7 +425,7 @@ def train(use_cuda, save_dirname=None, is_local=True): cost = cost[0] if batch_id % 10 == 0: - print("avg_cost:" + str(cost)) + print("avg_cost: " + str(cost)) if batch_id != 0: print("second per batch: " + str((time.time( ) - start_time) / batch_id)) diff --git a/07.label_semantic_roles/README.md b/07.label_semantic_roles/README.md index 36837cc3234d6770915b4c5d501058c3573dee9a..623d035da5ec71185f1269185f1dce8f366d2985 100644 --- a/07.label_semantic_roles/README.md +++ b/07.label_semantic_roles/README.md @@ -175,13 +175,6 @@ The raw data needs to be preprocessed into formats that PaddlePaddle can handle. 4. Construct the markings in BIO format; 5. Obtain the integer index corresponding to the word according to the dictionary. -```python -# import paddle.v2.dataset.conll05 as conll05 -# conll05.corpus_reader does step 1 and 2 as mentioned above. -# conll05.reader_creator does step 3 to 5. -# conll05.test gets preprocessed training instances. -``` - After preprocessing, a training sample contains nine features, namely: word sequence, predicate, predicate context (5 columns), region mark sequence, label sequence. The following table is an example of a training sample. | word sequence | predicate | predicate context（5 columns） | region mark sequence | label sequence| @@ -209,6 +202,8 @@ We trained a language model on the English Wikipedia to get a word vector lookup Here we fetch the dictionary, and print its size: ```python +from __future__ import print_function + import math, os import numpy as np import paddle @@ -223,9 +218,9 @@ word_dict_len = len(word_dict) label_dict_len = len(label_dict) pred_dict_len = len(verb_dict) -print word_dict_len -print label_dict_len -print pred_dict_len +print('word_dict_len: ', word_dict_len) +print('label_dict_len: ', label_dict_len) +print('pred_dict_len: ', pred_dict_len) ``` ## Model Configuration @@ -440,7 +435,7 @@ def train(use_cuda, save_dirname=None, is_local=True): cost = cost[0] if batch_id % 10 == 0: - print("avg_cost:" + str(cost)) + print("avg_cost: " + str(cost)) if batch_id != 0: print("second per batch: " + str((time.time( ) - start_time) / batch_id)) diff --git a/07.label_semantic_roles/index.cn.html b/07.label_semantic_roles/index.cn.html index 0a3276359a5f5deb40df536b249936d10fa4eba7..c19befb1b6a0a4ee679d52da838167c5f716b349 100644 --- a/07.label_semantic_roles/index.cn.html +++ b/07.label_semantic_roles/index.cn.html @@ -67,7 +67,7 @@ $$\mbox{[小明]}_{\mbox{Agent}}\mbox{[昨天]}_{\mbox{Time}}\mbox{[晚上]}_\mb 图1. 依存句法分析句法树示例 -然而，完全句法分析需要确定句子所包含的全部句法信息，并确定句子各成分之间的关系，是一个非常困难的任务，目前技术下的句法分析准确率并不高，句法分析的细微错误都会导致SRL的错误。为了降低问题的复杂度，同时获得一定的句法结构信息，“浅层句法分析”的思想应运而生。浅层句法分析也称为部分句法分析（partial parsing）或语块划分（chunking）。和完全句法分析得到一颗完整的句法树不同，浅层句法分析只需要识别句子中某些结构相对简单的独立成分，例如：动词短语，这些被识别出来的结构称为语块。为了回避 “无法获得准确率较高的句法树” 所带来的困难，一些研究\[[1](#参考文献)\]也提出了基于语块（chunk）的SRL方法。基于语块的SRL方法将SRL作为一个序列标注问题来解决。序列标注任务一般都会采用BIO表示方式来定义序列标注的标签集，我们先来介绍这种表示方法。在BIO表示法中，B代表语块的开始，I代表语块的中间，O代表语块结束。通过B、I、O 三种标记将不同的语块赋予不同的标签，例如：对于一个角色为A的论元，将它所包含的第一个语块赋予标签B-A，将它所包含的其它语块赋予标签I-A，不属于任何论元的语块赋予标签O。 +然而，完全句法分析需要确定句子所包含的全部句法信息，并确定句子各成分之间的关系，是一个非常困难的任务，目前技术下的句法分析准确率并不高，句法分析的细微错误都会导致SRL的错误。为了降低问题的复杂度，同时获得一定的句法结构信息，“浅层句法分析”的思想应运而生。浅层句法分析也称为部分句法分析（partial parsing）或语块划分（chunking）。和完全句法分析得到一颗完整的句法树不同，浅层句法分析只需要识别句子中某些结构相对简单的独立成分，例如：动词短语，这些被识别出来的结构称为语块。为了回避 “无法获得准确率较高的句法树” 所带来的困难，一些研究\[[1](#参考文献)\]也提出了基于语块（chunk）的SRL方法。基于语块的SRL方法将SRL作为一个序列标注问题来解决。序列标注任务一般都会采用BIO表示方式来定义序列标注的标签集，我们先来介绍这种表示方法。在BIO表示法中，B代表语块的开始，I代表语块的中间，O代表语块结束。通过B、I、O 三种标记将不同的语块赋予不同的标签，例如：对于一个由角色A拓展得到的语块组，将它所包含的第一个语块赋予标签B-A，将它所包含的其它语块赋予标签I-A，不属于任何论元的语块赋予标签O。 我们继续以上面的这句话为例，图1展示了BIO表示方法。 @@ -193,14 +193,6 @@ conll05st-release/ 4. 构造以BIO法表示的标记； 5. 依据词典获取词对应的整数索引。 - -```python -# import paddle.v2.dataset.conll05 as conll05 -# conll05.corpus_reader函数完成上面第1步和第2步. -# conll05.reader_creator函数完成上面第3步到第5步. -# conll05.test函数可以获取处理之后的每条样本来供PaddlePaddle训练. -``` - 预处理完成之后一条训练样本包含9个特征，分别是：句子序列、谓词、谓词上下文（占 5 列）、谓词上下区域标志、标注序列。下表是一条训练样本的示例。 | 句子序列 | 谓词 | 谓词上下文（窗口 = 5） | 谓词上下文区域标记 | 标注序列 | @@ -229,6 +221,8 @@ conll05st-release/ 获取词典，打印词典大小： ```python +from __future__ import print_function + import math, os import numpy as np import paddle @@ -243,9 +237,9 @@ word_dict_len = len(word_dict) label_dict_len = len(label_dict) pred_dict_len = len(verb_dict) -print word_dict_len -print label_dict_len -print pred_dict_len +print('word_dict_len: ', word_dict_len) +print('label_dict_len: ', label_dict_len) +print('pred_dict_len: ', pred_dict_len) ``` ## 模型配置说明 @@ -473,7 +467,7 @@ def train(use_cuda, save_dirname=None, is_local=True): cost = cost[0] if batch_id % 10 == 0: - print("avg_cost:" + str(cost)) + print("avg_cost: " + str(cost)) if batch_id != 0: print("second per batch: " + str((time.time( ) - start_time) / batch_id)) diff --git a/07.label_semantic_roles/index.html b/07.label_semantic_roles/index.html index 7d84612aa653746fdf247f2dbc44072d85937d0b..170c6b6f5916415bfbbb564680dc1efa12ef6de5 100644 --- a/07.label_semantic_roles/index.html +++ b/07.label_semantic_roles/index.html @@ -217,13 +217,6 @@ The raw data needs to be preprocessed into formats that PaddlePaddle can handle. 4. Construct the markings in BIO format; 5. Obtain the integer index corresponding to the word according to the dictionary. -```python -# import paddle.v2.dataset.conll05 as conll05 -# conll05.corpus_reader does step 1 and 2 as mentioned above. -# conll05.reader_creator does step 3 to 5. -# conll05.test gets preprocessed training instances. -``` - After preprocessing, a training sample contains nine features, namely: word sequence, predicate, predicate context (5 columns), region mark sequence, label sequence. The following table is an example of a training sample. | word sequence | predicate | predicate context（5 columns） | region mark sequence | label sequence| @@ -251,6 +244,8 @@ We trained a language model on the English Wikipedia to get a word vector lookup Here we fetch the dictionary, and print its size: ```python +from __future__ import print_function + import math, os import numpy as np import paddle @@ -265,9 +260,9 @@ word_dict_len = len(word_dict) label_dict_len = len(label_dict) pred_dict_len = len(verb_dict) -print word_dict_len -print label_dict_len -print pred_dict_len +print('word_dict_len: ', word_dict_len) +print('label_dict_len: ', label_dict_len) +print('pred_dict_len: ', pred_dict_len) ``` ## Model Configuration @@ -482,7 +477,7 @@ def train(use_cuda, save_dirname=None, is_local=True): cost = cost[0] if batch_id % 10 == 0: - print("avg_cost:" + str(cost)) + print("avg_cost: " + str(cost)) if batch_id != 0: print("second per batch: " + str((time.time( ) - start_time) / batch_id)) diff --git a/07.label_semantic_roles/train.py b/07.label_semantic_roles/train.py index f09ee33282aae551e746fa4458c7efd8a66cd991..780167a2cfa7b25d2d064efb700cac324cde3f6b 100644 --- a/07.label_semantic_roles/train.py +++ b/07.label_semantic_roles/train.py @@ -1,3 +1,5 @@ +from __future__ import print_function + import math, os import numpy as np import paddle diff --git a/08.machine_translation/README.cn.md b/08.machine_translation/README.cn.md index 014c9d79580bff9198bf26a0d38e03ac637dd84e..7457383100696b71cf3e8898a92c784dcfe030ee 100644 --- a/08.machine_translation/README.cn.md +++ b/08.machine_translation/README.cn.md @@ -85,7 +85,7 @@ 2. 将$z_{i+1}$通过`softmax`归一化，得到目标语言序列的第$i+1$个单词的概率分布$p_{i+1}$。概率分布公式如下： - $$p\left ( u_{i+1}|u_{<i+1},\mathbf{x} \right )=softmax(W_sz_{i+1}+b_z)$$ + $$p\left ( u_{i+1}|u_{