update to 0.0.9: Reduce memory allocation

README.md

@@ -48,7 +48,7 @@ print(words.words)
 print(words.postags) 
 #=> ['名詞', '助詞', '形状詞', '助動詞', '動詞', '名詞', '助動詞']
 
-# Nagisa gives you a simple word segmentation method.
+# The nagisa.wakati method is faster than the nagisa.tagging method.
 words = nagisa.wakati(text)
 print(words) 
 #=> ['Python', 'で', '簡単', 'に', '使える', 'ツール', 'です']

nagisa/model.py

@@ -4,7 +4,7 @@
 
 import numpy as np
 import dynet_config
-dynet_config.set(random_seed=1234)
+dynet_config.set(mem=32, random_seed=1234)
 import dynet as dy
 
 class Model(object):

setup.py

@@ -6,8 +6,14 @@
 from setuptools import setup
 from setuptools.extension import Extension
 
-with open('README.md') as f:
-    long_description = f.read()
+
+try:
+    from pypandoc import convert
+    long_description = convert('README.md', 'rst')
+except ImportError:
+    with open('README.md') as f:
+        long_description = f.read()
+
 
 classifiers = [
     'License :: OSI Approved :: MIT License',
@@ -20,6 +26,7 @@
     'Topic :: Software Development :: Libraries :: Python Modules'
 ]
 
+
 class defer_cythonize(list):
     def __init__(self, callback):
         self._list, self.callback = None, callback
@@ -52,11 +59,11 @@ def extensions():
     packages=['nagisa'],
     author = 'Taishi Ikeda',
     author_email = 'taishi.ikeda.0323@gmail.com',
-    version = '0.0.8',
-    description = 'Japanese word segmentation/POS tagging tool based on neural networks',
+    version = '0.0.9',
+    description = 'A Japanese tokenizer based on recurrent neural networks',
     long_description = long_description,
     url = 'https://github.com/taishi-i/nagisa',
-    download_url = 'https://github.com/taishi-i/nagisa/archive/0.0.8.tar.gz',
+    download_url = 'https://github.com/taishi-i/nagisa/archive/0.0.9.tar.gz',
     license = 'MIT License',
     platforms = 'Unix',
     setup_requires=['six', 'cython', 'numpy',],