TensorFlow DeepLab教程初稿-tensorflow gpu安装教程
Summary: DeepLab需要1.10以上版本。
本日志详细记录在两台不同笔记本电脑安装/更新 TensorFlow-GPU的具体过程
这是本人第3次，4次安装tf，这两次是gpu版。
第一次是安装cpu版，第二次是在python2.7 arcpy环境下安装32位 tf，但不能运行。第三次安装成功，但电脑配置过低，只能使用cpu功能；第四次安装成功，可以正确运行，使用gpu。
向日葵2019年7月24日于长沙
Confirm tensorflow installed correctly
>>> import tensorflow as tf
>>> hello=tf.constant("Hello tf")
>>> tf.Session().run(hello)
b'Hello tf'
查看tf版本
C:\Users\think>python --version
Python 3.6.7 :: Anaconda custom (64-bit)
Anaconda3 (3.6, 64-bit) interactive window [PTVS 16.0.19074.1-16.0]
Type $help for a list of commands.
>>> import tensorflow as tf
>>> tf.VERSION
'1.12.0'
>>>
>>> tf.__version__
'1.12.0'
查看CPU GPU版本
import os
from tensorflow.python.client import device_lib
os.environ["TF_CPP_MIN_LOG_LEVEL"]="99"
device_lib.list_local_devices()
[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 5642870862507944
]
>>> if __name__ == "main":
... print(device_lib.list_local_devices)
...
>>>
查看python安装路径
>>> import sys
>>> sys.path
['.', 'D:\\ProgramData\\Anaconda3\\', 'd:\\ProgramData\\Anaconda3\\python36.zip', 'd:\\ProgramData\\Anaconda3\\DLLs', 'd:\\ProgramData\\Anaconda3\\lib', 'd:\\ProgramData\\Anaconda3', 'd:\\ProgramData\\Anaconda3\\lib\\site-packages', 'd:\\ProgramData\\Anaconda3\\lib\\site-packages\\win32', 'd:\\ProgramData\\Anaconda3\\lib\\site-packages\\win32\\lib', 'd:\\ProgramData\\Anaconda3\\lib\\site-packages\\Pythonwin']
>>> import os
>>> os.path
<module 'ntpath' from 'd:\\ProgramData\\Anaconda3\\lib\\ntpath.py'>
Python.exe的路径：
"D:\ProgramData\Anaconda3\python.exe"
更新为gpu版本
Type cmd to enter dos command line and type:
pip3 install --ignore-installed --upgrade tensorflow-gpu
C:\Users\think>pip3 install --ignore-installed --upgrade tensorflow-gpu
Collecting tensorflow-gpu
Downloading https://files.pythonhosted.org/packages/c7/e8/f7ba3acc4e45bea553ef085846e0240daa71986a04e1819bafef569f055b/tensorflow_gpu-1.14.0-cp36-cp36m-win_amd64.whl (287.7MB)
由于空间原因，安装失败。
Installing collected packages: astor, wheel, wrapt, six, setuptools, protobuf, numpy, google-pasta, termcolor, grpcio, h5py, keras-applications, absl-py, keras-preprocessing, gast, tensorflow-estimator, markdown, werkzeug, tensorboard, tensorflow-gpu
Could not install packages due to an EnvironmentError: [Errno 28] No space left on device
重输入命令，可以续装，自动使用缓存。
tensorflow 1.12.0 has requirement tensorboard<1.13.0,>=1.12.0, but you'll have tensorboard 1.14.0 which is incompatible.
Installing collected packages: six, setuptools, protobuf, numpy, h5py, keras-applications, absl-py, grpcio, tensorflow-estimator, termcolor, google-pasta, wrapt, wheel, gast, werkzeug, markdown, tensorboard, astor, keras-preprocessing, tensorflow-gpu
Successfully installed absl-py-0.7.1 astor-0.8.0 gast-0.2.2 google-pasta-0.1.7 grpcio-1.22.0 h5py-2.9.0 keras-applications-1.0.8 keras-preprocessing-1.1.0 markdown-3.1.1 numpy-1.16.4 protobuf-3.9.0 setuptools-41.0.1 six-1.12.0 tensorboard-1.14.0 tensorflow-estimator-1.14.0 tensorflow-gpu-1.14.0 termcolor-1.1.0 werkzeug-0.15.5 wheel-0.33.4 wrapt-1.11.2
尝试卸载
C:\Users\think>pip3 uninstall tensorflow
Uninstalling tensorflow-1.12.0:
Would remove:
d:\programdata\anaconda3\lib\site-packages\tensorflow-1.12.0.dist-info\*
Proceed (y/n)? y
Successfully uninstalled tensorflow-1.12.0
成功卸载
重新尝试安装gpu版本：
C:\Users\think>pip3 install --ignore-installed --upgrade tensorflow-gpu
Collecting tensorflow-gpu
开始使用缓存安装
结果TensorFlow-gpu 安装成功
测试tf
C:\Users\think>python
Python 3.6.7 |Anaconda custom (64-bit)| (default, Oct 28 2018, 19:44:12) [MSC v.1915 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
D:\ProgramData\Anaconda3\lib\site-packages\numpy\core\__init__.py:29: UserWarning: loaded more than 1 DLL from .libs:
D:\ProgramData\Anaconda3\lib\site-packages\numpy\.libs\libopenblas.CSRRD7HKRKC3T3YXA7VY7TAZGLSWDKW6.gfortran-win_amd64.dll
D:\ProgramData\Anaconda3\lib\site-packages\numpy\.libs\libopenblas.TXA6YQSD3GCQQC22GEQ54J2UDCXDXHWN.gfortran-win_amd64.dll
stacklevel=1)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\__init__.py", line 28, in <module>
from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\__init__.py", line 47, in <module>
import numpy as np
File "D:\ProgramData\Anaconda3\lib\site-packages\numpy\__init__.py", line 142, in <module>
from . import core
File "D:\ProgramData\Anaconda3\lib\site-packages\numpy\core\__init__.py", line 91, in <module>
raise ImportError(msg.format(path))
ImportError: Something is wrong with the numpy installation. While importing we detected an older version of numpy in ['D:\\ProgramData\\Anaconda3\\lib\\site-packages\\numpy']. One method of fixing this is to repeatedly uninstall numpy until none is found, then reinstall this version.
Uninstall numpy
C:\Users\think>pip3 uninstall numpy
Uninstalling numpy-1.16.4:
Would remove:
d:\programdata\anaconda3\lib\site-packages\numpy-1.16.4.dist-info\*
d:\programdata\anaconda3\lib\site-packages\numpy\*
d:\programdata\anaconda3\scripts\f2py.exe
Would not remove (might be manually added):
d:\programdata\anaconda3\lib\site-packages\numpy\.libs\libopenblas.CSRRD7HKRKC3T3YXA7VY7TAZGLSWDKW6.gfortran-win_amd64.dll
d:\programdata\anaconda3\lib\site-packages\numpy\_import_tools.py
d:\programdata\anaconda3\lib\site-packages\numpy\_mklinit.cp36-win_amd64.pyd
d:\programdata\anaconda3\lib\site-packages\numpy\add_newdocs.py
d:\programdata\anaconda3\lib\site-packages\numpy\core\multiarray.cp36-win_amd64.pyd
d:\programdata\anaconda3\lib\site-packages\numpy\core\umath.cp36-win_amd64.pyd
d:\programdata\anaconda3\lib\site-packages\numpy\distutils\environment.py
d:\programdata\anaconda3\lib\site-packages\numpy\distutils\site.cfg
d:\programdata\anaconda3\lib\site-packages\numpy\random_intel\__init__.py
d:\programdata\anaconda3\lib\site-packages\numpy\random_intel\setup.py
d:\programdata\anaconda3\lib\site-packages\numpy\testing\_private\pytesttester.py
Proceed (y/n)? y
Successfully uninstalled numpy-1.16.4
Install numpy
C:\Users\think>pip3 install numpy
Requirement already satisfied: numpy in d:\programdata\anaconda3\lib\site-packages (1.15.4)
测试tf
>>> import tensorflow as tf
Traceback (most recent call last):
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\platform\self_check.py", line 75, in preload_check
ctypes.WinDLL(build_info.cudart_dll_name)
File "D:\ProgramData\Anaconda3\lib\ctypes\__init__.py", line 348, in __init__
self._handle = _dlopen(self._name, mode)
OSError: [WinError 126] The specified module could not be found
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\__init__.py", line 28, in <module>
from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\__init__.py", line 49, in <module>
from tensorflow.python import pywrap_tensorflow
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\pywrap_tensorflow.py", line 30, in <module>
self_check.preload_check()
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\platform\self_check.py", line 82, in preload_check
% (build_info.cudart_dll_name, build_info.cuda_version_number))
ImportError: Could not find 'cudart64_100.dll'. TensorFlow requires that this DLL be installed in a directory that is named in your %PATH% environment variable. Download and install CUDA 10.0 from this URL: https://developer.nvidia.com/cuda-90-download-archive（tf提示错误：应为https://developer.nvidia.com/cuda-100-download-archive）
结果：
2. tf安装程序的提示版本错误，但提示的dll版本信息是正确的。
由于已经安装了最新版的cuda，查看版本及文件名
G:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin
重新安装cuda v10.0
path
G:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\V10.0\bin
G:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\V10.0\libnvvp
G:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\V10.0
G:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\V10.0
Test tensorflow
>>> import tensorflow as tf
Traceback (most recent call last):
from tensorflow.python.pywrap_tensorflow_internal import *
_pywrap_tensorflow_internal = swig_import_helper()
_mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 243, in load_module
return load_dynamic(name, filename, file)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 343, in load_dynamic
ImportError: DLL load failed: The specified procedure could not be found.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\__init__.py", line 28, in <module>
from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
from tensorflow.python import pywrap_tensorflow
ImportError: Traceback (most recent call last):
from tensorflow.python.pywrap_tensorflow_internal import *
_pywrap_tensorflow_internal = swig_import_helper()
_mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 243, in load_module
return load_dynamic(name, filename, file)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 343, in load_dynamic
ImportError: DLL load failed: The specified procedure could not be found.
Failed to load the native TensorFlow runtime.
See https://www.tensorflow.org/install/errors
for some common reasons and solutions. Include the entire stack trace
above this error message when asking for help.
Reinstall cudnn
官网下载好后可以直接解压文件夹。然后将这个文件夹下的文件按照如下操作复制到CUDA路径下：
测试：
>>> import tensorflow as tf
Traceback (most recent call last):
from tensorflow.python.pywrap_tensorflow_internal import *
_pywrap_tensorflow_internal = swig_import_helper()
_mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 243, in load_module
return load_dynamic(name, filename, file)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 343, in load_dynamic
ImportError: DLL load failed: The specified procedure could not be found.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\__init__.py", line 28, in <module>
from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
from tensorflow.python import pywrap_tensorflow
ImportError: Traceback (most recent call last):
from tensorflow.python.pywrap_tensorflow_internal import *
_pywrap_tensorflow_internal = swig_import_helper()
_mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 243, in load_module
return load_dynamic(name, filename, file)
File "D:\ProgramData\Anaconda3\lib\imp.py", line 343, in load_dynamic
ImportError: DLL load failed: The specified procedure could not be found.
Failed to load the native TensorFlow runtime.
See https://www.tensorflow.org/install/errors
for some common reasons and solutions. Include the entire stack trace
above this error message when asking for help.
结果：依然出错
卸载
C:\Users\think>pip3 uninstall tensorflow-gpu
Uninstalling tensorflow-gpu-1.14.0:
d:\programdata\anaconda3\lib\site-packages\tensorflow\*
d:\programdata\anaconda3\lib\site-packages\tensorflow_gpu-1.14.0.dist-info\*
d:\programdata\anaconda3\scripts\freeze_graph.exe
d:\programdata\anaconda3\scripts\saved_model_cli.exe
d:\programdata\anaconda3\scripts\tensorboard.exe
d:\programdata\anaconda3\scripts\tf_upgrade_v2.exe
d:\programdata\anaconda3\scripts\tflite_convert.exe
d:\programdata\anaconda3\scripts\toco.exe
d:\programdata\anaconda3\scripts\toco_from_protos.exe
Run python3.7 installed with vs
C:\Program Files (x86)\Microsoft Visual Studio\Shared\Python37_64
C:\Program Files (x86)\Microsoft Visual Studio\Shared\Python37_64\Scripts
重装
Collecting tensorflow-gpu
Installing collected packages: tensorflow-estimator, google-pasta, numpy, six, grpcio, astor, wheel, wrapt, gast, absl-py, setuptools, werkzeug, protobuf, markdown, tensorboard, h5py, keras-applications, keras-preprocessing, termcolor, tensorflow-gpu
Could not install packages due to an EnvironmentError: [WinError 5] Access is denied: 'c:\\program files (x86)\\microsoft visual studio\\shared\\python37_64\\Lib\\site-packages\\tensorflow_estimator'
Consider using the `--user` option or check the permissions.
You are using pip version 18.1, however version 19.2.1 is available.
You should consider upgrading via the 'python -m pip install --upgrade pip' command.
更新pip
结果：失败
100% |████████████████████████████████| 1.4MB 6.0MB/s
Installing collected packages: pip
Found existing installation: pip 18.1
Consider using the `--user` option or check the permissions.
You are using pip version 18.1, however version 19.2.1 is available.
You should consider upgrading via the 'python -m pip install --upgrade pip' command.
设置所有权限
重新更新pip
结果：成功
Installing collected packages: pip
Found existing installation: pip 18.1
Successfully uninstalled pip-18.1
Successfully installed pip-19.2.1
重新安装tf-gpu
Collecting tensorflow-gpu
Using cached https://files.pythonhosted.org/packages/81/d1/9222b9aac2fa27dccaef38917cde84c24888f3cd0dd139c7e12be9f49a7a/tensorflow_gpu-1.14.0-cp37-cp37m-win_amd64.whl
WARNING: The script f2py.exe is installed in 'c:\program files (x86)\microsoft visual studio\shared\python37_64\Scripts' which is not on PATH.
Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.
Running setup.py install for wrapt ... done
Found existing installation: setuptools 40.6.2
Uninstalling setuptools-40.6.2:
Successfully uninstalled setuptools-40.6.2
WARNING: The script wheel.exe is installed in 'c:\program files (x86)\microsoft visual studio\shared\python37_64\Scripts' which is not on PATH.
Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.
Running setup.py install for absl-py ... done
WARNING: The script markdown_py.exe is installed in 'c:\program files (x86)\microsoft visual studio\shared\python37_64\Scripts' which is not on PATH.
Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.
WARNING: The script tensorboard.exe is installed in 'c:\program files (x86)\microsoft visual studio\shared\python37_64\Scripts' which is not on PATH.
Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.
Running setup.py install for termcolor ... done
Running setup.py install for gast ... done
WARNING: The scripts freeze_graph.exe, saved_model_cli.exe, tensorboard.exe, tf_upgrade_v2.exe, tflite_convert.exe, toco.exe and toco_from_protos.exe are installed in 'c:\program files (x86)\microsoft visual studio\shared\python37_64\Scripts' which is not on PATH.
Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.
Successfully installed absl-py-0.7.1 astor-0.8.0 gast-0.2.2 google-pasta-0.1.7 grpcio-1.22.0 h5py-2.9.0 keras-applications-1.0.8 keras-preprocessing-1.1.0 markdown-3.1.1 numpy-1.16.4 protobuf-3.9.0 setuptools-41.0.1 six-1.12.0 tensorboard-1.14.0 tensorflow-estimator-1.14.0 tensorflow-gpu-1.14.0 termcolor-1.1.0 werkzeug-0.15.5 wheel-0.33.4 wrapt-1.11.2
添加path环境变量
c:\program files (x86)\microsoft visual studio\shared\python37_64\Scripts
测试tf
>>> import tensorflow as tf
>>> tf.__version__
'1.14.0'
结果：成功
测试cpu还是gpu
>>> import os
>>> from tensorflow.python.client import device_lib
>>> os.environ["TF_CPP_MIN_LOG_LEVEL"]="99"
>>> device_lib.list_local_devices()
[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 11073789457534317226
]
测试是否在gpu上运行
>>> import tensorflow as tf
>>> sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
2019-07-24 18:55:42.319137: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2019-07-24 18:55:42.324349: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library nvcuda.dll
2019-07-24 18:55:42.859900: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties:
name: Quadro K1100M major: 3 minor: 0 memoryClockRate(GHz): 0.7055
pciBusID: 0000:01:00.0
2019-07-24 18:55:42.864411: I tensorflow/stream_executor/platform/default/dlopen_checker_stub.cc:25] GPU libraries are statically linked, skip dlopen check.
2019-07-24 18:55:42.869847: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1717] Ignoring visible gpu device (device: 0, name: Quadro K1100M, pci bus id: 0000:01:00.0, compute capability: 3.0) with Cuda compute capability 3.0. The minimum required Cuda capability is 3.5.
2019-07-24 18:55:42.902715: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-07-24 18:55:42.905383: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187] 0
2019-07-24 18:55:42.907013: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0: N
Device mapping: no known devices.
2019-07-24 18:55:42.910171: I tensorflow/core/common_runtime/direct_session.cc:296] Device mapping:
结果：
由于tf 最小cuda计算能力要求为3.5，本gpu(device: 0, name: Quadro K1100M, pci bus id: 0000:01:00.0, compute capability: 3.0)不满足要求。
换台电脑，在python3.6上安装tf-gpu
主要内容：
1. 添加权限
2 安装cuda，cudnn
3 根据提示更新pip
4 安装tf-gpu
测试是否在gpu上运行
>>> import tensorflow as tf
>>> sess=tf.Session(config=tf.ConfigProto(log_device_placement=True))
2019-07-24 19:05:52.174361: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2019-07-24 19:05:52.180939: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library nvcuda.dll
2019-07-24 19:05:52.309641: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1640] Found device 0 with properties:
name: GeForce 940MX major: 5 minor: 0 memoryClockRate(GHz): 1.2415
pciBusID: 0000:01:00.0
2019-07-24 19:05:52.317307: I tensorflow/stream_executor/platform/default/dlopen_checker_stub.cc:25] GPU libraries are statically linked, skip dlopen check.
2019-07-24 19:05:52.322696: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1763] Adding visible gpu devices: 0
2019-07-24 19:05:53.599442: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1181] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-07-24 19:05:53.603636: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1187] 0
2019-07-24 19:05:53.605884: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1200] 0: N
2019-07-24 19:05:53.608481: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1326] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 1391 MB memory) -> physical GPU (device: 0, name: GeForce 940MX, pci bus id: 0000:01:00.0, compute capability: 5.0)
Device mapping:
/job:localhost/replica:0/task:0/device:GPU:0 -> device: 0, name: GeForce 940MX, pci bus id: 0000:01:00.0, compute capability: 5.0
2019-07-24 19:05:53.618489: I tensorflow/core/common_runtime/direct_session.cc:296] Device mapping:
/job:localhost/replica:0/task:0/device:GPU:0 -> device: 0, name: GeForce 940MX, pci bus id: 0000:01:00.0, compute capability: 5.0
进一步测试
>>> import os
>>> from tensorflow.python.client import device_lib
>>> os.environ["TF_CPP_MIN_LOG_LEVEL"] = "99"
>>> print(device_lib.list_local_devices())
[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 2949566760586279955
,
name: "/device:GPU:0"
device_type: "GPU"
memory_limit: 1459018137
locality {
bus_id: 1
links {
}
}
incarnation: 14182985404273149922
physical_device_desc: "device: 0, name: GeForce 940MX, pci bus id: 0000:01:00.0, compute capability: 5.0"
]
总结
在anaconda3.6安装最新版（未指定版本）tf-gpu，失败。因为没有提示具体错误，难以分析具体原因，所以拟在别的位置的python上重新安装。
失败的原因可能是在本python下已经安装了tf cpu '1.12.0'
找到vs安装的python3.7，添加环境变量和权限后，更新pip命令，未指定版本， '1.14.0'，安装成功，发现下载速度很快，一会儿就安装好了，但不能正确运行。
之后换位置安装好，但gpu版本太低，无法应用gpu加速。
最后换电脑安装，终于成功。
结论：
1. 查看gpu-cuda计算能力，是否大于3.5
2. 如果python安装c盘，需要添加权限
3 更新pip
4. 安装tf-gpu，目前最新版为1.14.0
5.安装cuda和cudnn，需要完全核定版本，这一步可以根据运行tf的结果来确认版本。
