2017-04-03 108 views
0

我有Ubuntu 16.04服务器环境,并试图在其上安装CNTK。虽然我试图在环境部分安装pip安装,但出现以下错误。在安装.whl文件时安装CNTK - MemoryError时出错

我成功地低于2步跑:

$畅达创建--name cntk-PY34蟒蛇= 3.4 numpy的SciPy的h5py jupyter

$激活cntk-py35

但是,当我尝试安装该cntk WHL文件我得到一个错误:

$ PIP安装https://cntk.ai/PythonWheel/CPU-Only/cntk-2.0.beta15.0-cp35-cp35m-linux_x86_64.whl

========错误==================

例外: 回溯(最近通话最后一个): 文件 “/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/basecommand.py”,行215,主 status = self.run(options,args) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/commands/install.py” ,第335行,运行 wb.build(autobuilding = True) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/wheel.py”,行749,in build self.requirement_set.prepare_files(self.finder) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/req/req_set.py” ,380行,在prepare_files ignore_d ependencies = self.ignore_dependencies)) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/req/req_set.py”,第620行,在_prepare_file中 会话= self.session,hashes = hashes) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/download.py”,第821行,在unpack_url中 散列=散列 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/download.py”,行659,in unpack_http_url 哈希) 文件“/ home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/download.py“,行853,位于_download_http_url stream = True, File”/home/ubuntu/.conda/envs /cntk-py35/lib/python3.5/site-packages/pip/_vendor/requests/sessions.py“,第488行,获得 return self.request('GET',url,** kwargs) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/download.py”,第386行,请求 return super(PipSession,self).request(method,url,* args,** kwargs) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/ site-packages/pip/_vendor/requests/sessions.py“,第475行,请求 resp = self.send(prep,** send_kwargs) 文件”/home/ubuntu/.conda/envs/cntk-py35/ lib/python3.5/site-packages/pip/_vendor/requests/sessions.py“,第596行,发送 r = adapter.send(request,** kwargs) 文件”/home/ubuntu/.conda/ envs/cntk-py35/lib/python3.5/site-packages/pip/_vendor/cachecontrol/adapter.py“,第37行,发送 cached_response = self.controller.cached_request(request ) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/_vendor/cachecontrol/controller.py”,第111行,在cached_request中 resp = self.serializer .loads(request,cache_data) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3.5/site-packages/pip/_vendor/cachecontrol/serialize.py”,第114行,载入 return getattr(self,“_loads_v {0}”。format(ver))(request,data) 文件“/home/ubuntu/.conda/envs/cntk-py35/lib/python3。5 /站点包/ PIP/_vendor/cachecontrol/serialize.py”,线路176,在_loads_v2 缓存= json.loads(zlib.decompress(数据).decode( “UTF8”)) 的MemoryError

不限想法?

提前感谢!

回答

0

在你的问题,你提到的两个不同版本的Python(3.4和3.5)。此外,蟒蛇环境激活需要进行采购。假设你已经满足的openmpi依赖(参考https://github.com/Microsoft/CNTK/wiki/Setup-Linux-Python),你可以尝试其中的一种:

# For a Python 3.4 based setup 
conda create --name cntk-py34 python=3.4 numpy scipy h5py jupyter 
source activate cntk-py34 
pip install https://cntk.ai/PythonWheel/CPU-Only/cntk-2.0rc1-cp34-cp34m-linux_x86_64.whl 

# For a Python 3.5 based setup 
conda create --name cntk-py35 python=3.5 numpy scipy h5py jupyter 
source activate cntk-py35 
pip install https://cntk.ai/PythonWheel/CPU-Only/cntk-2.0rc1-cp35-cp35m-linux_x86_64.whl