在我们 使用tensorflow 和 spark 结合的时候 ,肯定非常激动,关键 我们打算 使用哪一种语言 建构我们的机器学习代码,最主要的的有四种 ,python java scala R,当然 python 是门槛较低的。使用java scala 一般人不一定能hold 的住,所以我们首先 讲 python版的工具链
首先 假设你已经有了一台 mac pro
安装了 python 3.5 或者3.6
jdk 8 ,最好不要用jdk9 ,jdk9 有很多问题
本地 安装 了 tensorflow 和pyspark
当然 本地 homebrew 安装 spark 2.1 和hadoop 2.8.1
想要 spark 和tensorflow 串联在一起,你需要 spark-deeplearning 这个神器
我们git 一下 威廉大哥的 ,因为 威廉大哥扩展了 原来 的databircks ,添加了新功能的,一定要用威廉大哥的,否则 威廉大哥用的一些 方法 你找不到的,比如
from sparkdl import TextEstimator
from sparkdl.transformers.easy_feature import EasyFeature
这些都是威廉自己扩展的,非常好用
https://github.com/allwefantasy/spark-deep-learning
mkdir sparkdl && cd sparkdl
git clone https://github.com/allwefantasy/spark-deep-learning.git . # 注意 点号
git checkout release
当然 单单它还不够,我们还需要 tensorflow 与spark 底层元素交互的 媒介
tensorframe,并且 spark-DeepLearning 本身依赖 tensorframes
https://github.com/databricks/tensorframes/
git clone https://github.com/databricks/tensorframes.git
本身 如果你想直接使用pip 来安装这两个包 ,抱歉 pip仓库没有。
这里就涉及到了 pip 安装 本地 package
其实 也没有多难,幸好威廉大哥给了我一些锦囊妙计
想pip 安装本地的package ,大致分两步,
1.创建 一个启动文件 setup.py, setup 文件可以参考威廉大哥的
[https://github.com/allwefantasy/spark-deep-learning/blob/release/python/setup.py](https://github.com/allwefantasy/spark-deep-learning/blob/release/python/setup.py)
2.在 setup.py文件中 配置package的属性 文件路径 等信息
- 执行 一系列命令 最重要的是 执行
python setup.py bdist_wheel
这样就会生成 二进制文件 package Name-Version-py3.whl
4.然后 进入这个文件目录 在Terminal 中执行
pip install package-Name-Version-py3.whl # [anaconda]
pip3 install package-Name-Version-py3.whl # [python 3.6]
- pip list 和 pip3 list 验证 是否安装成功
- import package 查看模块是否真实可以使用
这里面要注意的就是 在执行 2的时候 要确定在同目录下必须有package对应的python源文件,否则即使生成了 whl文件,这个包也是一个假的不可以被使用的
另外 编写 setup.py文件 要注意的就是包名和版本要 与实际一致,否则可能真包安装了,本身依赖的其他包也找不到它
sparkdl 的 setup.py
import codecs
import os
from setuptools import setup, find_packages
# See this web page for explanations:
# https://hynek.me/articles/sharing-your-labor-of-love-pypi-quick-and-dirty/
PACKAGES = ["sparkdl"]
KEYWORDS = ["spark", "deep learning", "distributed computing", "machine learning"]
CLASSIFIERS = [
"Programming Language :: Python :: 2.7",
"Programming Language :: Python :: 3.4",
"Programming Language :: Python :: 3.5",
"Development Status :: 4 - Beta",
"Intended Audience :: Developers",
"Natural Language :: English",
"License :: OSI Approved :: Apache Software License",
"Operating System :: OS Independent",
"Programming Language :: Python",
"Topic :: Scientific/Engineering",
]
# Project root
ROOT = os.path.abspath(os.path.dirname(__file__))
#
#
# def read(*parts):
# """
# Build an absolute path from *parts* and and return the contents of the
# resulting file. Assume UTF-8 encoding.
# """
# with codecs.open(os.path.join(ROOT, *parts), "rb", "utf-8") as f:
# return f.read()
#
# def configuration(parent_package='', top_path=None):
# if os.path.exists('MANIFEST'):
# os.remove('MANIFEST')
#
# from numpy.distutils.misc_util import Configuration
# config = Configuration(None, parent_package, top_path)
#
# # Avoid non-useful msg:
# # "Ignoring attempt to set 'name' (from ... "
# config.set_options(ignore_setup_xxx_py=True,
# assume_default_configuration=True,
# delegate_options_to_subpackages=True,
# quiet=True)
#
# config.add_subpackage('sparkdl')
#
# return config
setup(
name="sparkdl",
description="Integration tools for running deep learning on Spark",
license="Apache 2.0",
url="https://github.com/allwefantasy/spark-deep-learning",
version="0.2.2",
author="Joseph Bradley",
author_email="joseph@databricks.com",
maintainer="Tim Hunter",
maintainer_email="timhunter@databricks.com",
keywords=KEYWORDS,
packages=find_packages(),
classifiers=CLASSIFIERS,
zip_safe=False,
include_package_data=True
)
tensorframes 的 setup.py
import codecs
import os
from setuptools import setup, find_packages
# See this web page for explanations:
# https://hynek.me/articles/sharing-your-labor-of-love-pypi-quick-and-dirty/
PACKAGES = ["tensorframes"]
KEYWORDS = ["spark", "deep learning", "distributed computing", "machine learning"]
CLASSIFIERS = [
"Programming Language :: Python :: 2.7",
"Programming Language :: Python :: 3.4",
"Programming Language :: Python :: 3.5",
"Development Status :: 4 - Beta",
"Intended Audience :: Developers",
"Natural Language :: English",
"License :: OSI Approved :: Apache Software License",
"Operating System :: OS Independent",
"Programming Language :: Python",
"Topic :: Scientific/Engineering",
]
# Project root
ROOT = os.path.abspath(os.path.dirname(__file__))
#
#
# def read(*parts):
# """
# Build an absolute path from *parts* and and return the contents of the
# resulting file. Assume UTF-8 encoding.
# """
# with codecs.open(os.path.join(ROOT, *parts), "rb", "utf-8") as f:
# return f.read()
#
# def configuration(parent_package='', top_path=None):
# if os.path.exists('MANIFEST'):
# os.remove('MANIFEST')
#
# from numpy.distutils.misc_util import Configuration
# config = Configuration(None, parent_package, top_path)
#
# # Avoid non-useful msg:
# # "Ignoring attempt to set 'name' (from ... "
# config.set_options(ignore_setup_xxx_py=True,
# assume_default_configuration=True,
# delegate_options_to_subpackages=True,
# quiet=True)
#
# config.add_subpackage('sparkdl')
#
# return config
setup(
name="tensorframes",
description="Integration tools for running deep learning on Spark",
license="Apache 2.0",
url="https://github.com/databricks/tensorframes",
version="0.2.9",
author="Joseph Bradley",
author_email="joseph@databricks.com",
maintainer="Tim Hunter",
maintainer_email="timhunter@databricks.com",
keywords=KEYWORDS,
packages=find_packages(),
classifiers=CLASSIFIERS,
zip_safe=False,
include_package_data=True
)
然后 先 安装 sparkdl ,进入 spark-deeplearning 目录 打开Terminal
cd ./python && python setup.py bdist_wheel && cd dist
pip install sparkdl-0.2.2-py3-none-any.whl
pip3 install sparkdl-0.2.2-py3-none-any.whl
在安装tensorframes 要注意的就是 tensorframes的根目录下的python目录没有对应的源文件,需要找到源文件 复制到这里,一定要把 ./src/main/python/ 下的两个文件目录 tensorframes 和tensorframes_snippets 拷贝到 根目录下的python目录下,否则即使 安装了tensorframes 也是不可以用的,另外一定要 提前把pyspark 安装好,否则 也是不可以用的
cp ./src/main/python/* ./python
cd ./python && python setup.py bdist_wheel && cd dist
pip install tensorframes-0.2.9-py3-none-any.whl
pip3 install tensorframes-0.2.9-py3-none-any.whl
命令可以参考 威廉大哥的
cd ./python && python [setup.py](setup.py) bdist_wheel && cd dist && pip uninstall sparkdl && pip install ./sparkdl-0.2.2-py2-none-any.whl && cd ..
然后我们在pycharm里就可以愉快的使用了