update marddown, use numpy.savetext and numpy.loadtxt to save/load embedding table

04.word2vec/README.cn.md

@@ -207,6 +207,26 @@ hiddensize = 256 # 隐层维度
 N = 5 # 训练5-Gram
 ```
 
+用于保存和加载word_dict和embedding table的函数
+```python
+def wordemb(inlayer):
+    wordemb = paddle.layer.table_projection(
+        input=inlayer,
+        size=embsize,
+        param_attr=paddle.attr.Param(
+            name="_proj", initial_std=0.001, learning_rate=1, l2_rate=0))
+    return wordemb
+
+
+# save and load word dict and embedding table
+def save_dict_and_embedding(word_dict, embeddings):
+    with open("word_dict", "w") as f:
+        for key in word_dict:
+            f.write(key + " " + str(word_dict[key]) + "\n")
+    with open("embedding_table", "w") as f:
+        numpy.savetxt(f, embeddings, delimiter=',', newline='\n')
+```
+
 接着，定义网络结构：
 
 - 将$w_t$之前的$n-1$个词 $w_{t-n+1},...w_{t-1}$，通过$|V|\times D$的矩阵映射到D维词向量（本例中取D=32）。
@@ -333,6 +353,16 @@ Pass 0, Batch 200, Cost 5.786797, {'classification_error_evaluator': 0.8125}, Te
 
 经过30个pass，我们将得到平均错误率为classification_error_evaluator=0.735611。
 
+## 保存词典和embedding
+
+训练完成之后，我们可以把词典和embedding table单独保存下来，后面可以直接使用
+
+```python
+# save word dict and embedding table
+embeddings = parameters.get("_proj").reshape(len(word_dict), embsize)
+save_dict_and_embedding(word_dict, embeddings)
+```
+
 
 ## 应用模型
 训练模型后，我们可以加载模型参数，用训练出来的词向量初始化其他模型，也可以将模型查看参数用来做后续应用。

04.word2vec/README.md

@@ -224,6 +224,27 @@ hiddensize = 256 # hidden layer dimension
 N = 5 # train 5-gram
 ```
 
+
+- functions used to save and load word dict and embedding table
+```python
+def wordemb(inlayer):
+    wordemb = paddle.layer.table_projection(
+        input=inlayer,
+        size=embsize,
+        param_attr=paddle.attr.Param(
+            name="_proj", initial_std=0.001, learning_rate=1, l2_rate=0))
+    return wordemb
+
+
+# save and load word dict and embedding table
+def save_dict_and_embedding(word_dict, embeddings):
+    with open("word_dict", "w") as f:
+        for key in word_dict:
+            f.write(key + " " + str(word_dict[key]) + "\n")
+    with open("embedding_table", "w") as f:
+        numpy.savetxt(f, embeddings, delimiter=',', newline='\n')
+```
+
 - Map the $n-1$ words $w_{t-n+1},...w_{t-1}$ before $w_t$ to a D-dimensional vector though matrix of dimention $|V|\times D$ (D=32 in this example).
 
 ```python
@@ -343,6 +364,16 @@ Pass 0, Batch 200, Cost 5.786797, {'classification_error_evaluator': 0.8125}, Te
 
 After 30 passes, we can get average error rate around 0.735611.
 
+## Save word dict and embedding table
+
+after training, we can save the word dict and embedding table for the future usage.
+
+```python
+# save word dict and embedding table
+embeddings = parameters.get("_proj").reshape(len(word_dict), embsize)
+save_dict_and_embedding(word_dict, embeddings)
+```
+
 
 ## Model Application
 

04.word2vec/index.cn.html

@@ -249,6 +249,26 @@ hiddensize = 256 # 隐层维度
 N = 5 # 训练5-Gram
 ```
 
+用于保存和加载word_dict和embedding table的函数
+```python
+def wordemb(inlayer):
+    wordemb = paddle.layer.table_projection(
+        input=inlayer,
+        size=embsize,
+        param_attr=paddle.attr.Param(
+            name="_proj", initial_std=0.001, learning_rate=1, l2_rate=0))
+    return wordemb
+
+
+# save and load word dict and embedding table
+def save_dict_and_embedding(word_dict, embeddings):
+    with open("word_dict", "w") as f:
+        for key in word_dict:
+            f.write(key + " " + str(word_dict[key]) + "\n")
+    with open("embedding_table", "w") as f:
+        numpy.savetxt(f, embeddings, delimiter=',', newline='\n')
+```
+
 接着，定义网络结构：
 
 - 将$w_t$之前的$n-1$个词 $w_{t-n+1},...w_{t-1}$，通过$|V|\times D$的矩阵映射到D维词向量（本例中取D=32）。
@@ -375,6 +395,16 @@ Pass 0, Batch 200, Cost 5.786797, {'classification_error_evaluator': 0.8125}, Te
 
 经过30个pass，我们将得到平均错误率为classification_error_evaluator=0.735611。
 
+## 保存词典和embedding
+
+训练完成之后，我们可以把词典和embedding table单独保存下来，后面可以直接使用
+
+```python
+# save word dict and embedding table
+embeddings = parameters.get("_proj").reshape(len(word_dict), embsize)
+save_dict_and_embedding(word_dict, embeddings)
+```
+
 
 ## 应用模型
 训练模型后，我们可以加载模型参数，用训练出来的词向量初始化其他模型，也可以将模型查看参数用来做后续应用。

04.word2vec/index.html

@@ -266,6 +266,27 @@ hiddensize = 256 # hidden layer dimension
 N = 5 # train 5-gram
 ```
 
+
+- functions used to save and load word dict and embedding table
+```python
+def wordemb(inlayer):
+    wordemb = paddle.layer.table_projection(
+        input=inlayer,
+        size=embsize,
+        param_attr=paddle.attr.Param(
+            name="_proj", initial_std=0.001, learning_rate=1, l2_rate=0))
+    return wordemb
+
+
+# save and load word dict and embedding table
+def save_dict_and_embedding(word_dict, embeddings):
+    with open("word_dict", "w") as f:
+        for key in word_dict:
+            f.write(key + " " + str(word_dict[key]) + "\n")
+    with open("embedding_table", "w") as f:
+        numpy.savetxt(f, embeddings, delimiter=',', newline='\n')
+```
+
 - Map the $n-1$ words $w_{t-n+1},...w_{t-1}$ before $w_t$ to a D-dimensional vector though matrix of dimention $|V|\times D$ (D=32 in this example).
 
 ```python
@@ -385,6 +406,16 @@ Pass 0, Batch 200, Cost 5.786797, {'classification_error_evaluator': 0.8125}, Te
 
 After 30 passes, we can get average error rate around 0.735611.
 
+## Save word dict and embedding table
+
+after training, we can save the word dict and embedding table for the future usage.
+
+```python
+# save word dict and embedding table
+embeddings = parameters.get("_proj").reshape(len(word_dict), embsize)
+save_dict_and_embedding(word_dict, embeddings)
+```
+
 
 ## Model Application
 

04.word2vec/train.py

-import math, os
-import numpy
+import math
+import os
 
+import numpy
 import paddle.v2 as paddle
 
 with_gpu = os.getenv('WITH_GPU', '0') != '0'
@@ -25,22 +26,17 @@ def save_dict_and_embedding(word_dict, embeddings):
         for key in word_dict:
             f.write(key + " " + str(word_dict[key]) + "\n")
     with open("embedding_table", "w") as f:
-        for line in embeddings:
-            f.write(",".join([str(x) for x in line]) + "\n")
+        numpy.savetxt(f, embeddings, delimiter=',', newline='\n')
 
 
 def load_dict_and_embedding():
     word_dict = dict()
-    embeddings = []
-
     with open("word_dict", "r") as f:
         for line in f:
             key, value = line.strip().split(" ")
             word_dict[key] = value
-    with open("embedding_table", "r") as f:
-        for line in f:
-            embeddings.append(
-                numpy.array([float(x) for x in line.strip().split(',')]))
+
+    embeddings = numpy.loadtxt("embedding_table", delimiter=",")
     return word_dict, embeddings
 
 
@@ -102,7 +98,7 @@ def main():
     trainer = paddle.trainer.SGD(cost, parameters, adagrad)
     trainer.train(
         paddle.batch(paddle.dataset.imikolov.train(word_dict, N), 32),
-        num_passes=1,
+        num_passes=100,
         event_handler=event_handler)
 
     # save word dict and embedding table