win10安装tensorflow-gpu-1.15
因课程要求，需要安装tensorflow1.15。我是使用conda来管理不同环境的。
总的步骤为：
- 查看tensorflow1.15对应python, cuda, cudnn版本
- 安装python, cuda, cudnn
- 安装tensorflow
查看对应版本
查看tensorflow1.15对应python版本：https://tensorflow.google.cn/install/source#gpu
可以看出我需要python3.7, cuda10.0, cudnn7.4
安装python, cuda, cudnn
conda创建名为tf1的python3.7环境
conda create --prefix=D:\Anaconda3\envs\tf1 python=3.7
进入tf1环境
conda activate tf1
在conda源中搜索cudatoolkit
conda search cudatoolkit --info
找到对应cudatoolkit
复制其url并下载文件
在conda源中搜索cudnn
conda search cudnn --info
没有看到cudnn7.4，但是这个cudnn7.6也与cuda10.0相容的样子，于是下载这个（事实证明，确实可以）
复制其url并下载文件
在tf1环境中安装上面两个文件
conda install --use-local E:\study\deapLearning\cuda\cudatoolkit-10.0.130-0.conda
conda install --use-local E:\study\deapLearning\cuda\cudnn-7.6.0-cuda10.0_0.conda
安装好后，执行conda list命令可以看到这两个的信息，说明安装成功
安装tensorflow1.15
可以查阅官网 https://tensorflow.google.cn/install/gpu
使用pip安装gpu版tensorflow1.15，注意，这里我忘了换pip源，导致tensorflow下载了好几个小时，读者在这里记得换源
pip install tensorflow-gpu==1.15
安装好后，打开python试一试：
import tensorflow as tf
出现错误：
>>> import tensorflow as tf
2022-11-25 23:28:12.772433: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow\__init__.py", line 99, in <module>
from tensorflow_core import *
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\__init__.py", line 28, in <module>
from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
File "<frozen importlib._bootstrap>", line 1019, in _handle_fromlist
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow\__init__.py", line 50, in __getattr__
module = self._load()
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow\__init__.py", line 44, in _load
module = _importlib.import_module(self.__name__)
File "D:\Anaconda3\envs\tf1\lib\importlib\__init__.py", line 127, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\python\__init__.py", line 52, in <module>
from tensorflow.core.framework.graph_pb2 import *
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\core\framework\graph_pb2.py", line 16, in <module>
from tensorflow.core.framework import node_def_pb2 as tensorflow_dot_core_dot_framework_dot_node__def__pb2
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\core\framework\node_def_pb2.py", line 16, in <module>
from tensorflow.core.framework import attr_value_pb2 as tensorflow_dot_core_dot_framework_dot_attr__value__pb2
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\core\framework\attr_value_pb2.py", line 16, in <module>
from tensorflow.core.framework import tensor_pb2 as tensorflow_dot_core_dot_framework_dot_tensor__pb2
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\core\framework\tensor_pb2.py", line 16, in <module>
from tensorflow.core.framework import resource_handle_pb2 as tensorflow_dot_core_dot_framework_dot_resource__handle__pb2
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\core\framework\resource_handle_pb2.py", line 16, in <module>
from tensorflow.core.framework import tensor_shape_pb2 as tensorflow_dot_core_dot_framework_dot_tensor__shape__pb2
File "D:\Anaconda3\envs\tf1\lib\site-packages\tensorflow_core\core\framework\tensor_shape_pb2.py", line 42, in <module>
serialized_options=None, file=DESCRIPTOR),
File "D:\Anaconda3\envs\tf1\lib\site-packages\google\protobuf\descriptor.py", line 560, in __new__
_message.Message._CheckCalledFromGeneratedFile()
TypeError: Descriptors cannot not be created directly.
If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.
If you cannot immediately regenerate your protos, some other possible workarounds are:
1. Downgrade the protobuf package to 3.20.x or lower.
2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).
More information: https://developers.google.com/protocol-buffers/docs/news/2022-05-06#python-updates
>>>
这是由于我们用pip安装tensorflow的时候，安装依赖包protobuf的版本过高造成的。新装的pytorch也会有这样的问题。[1]
卸掉原来的，再重新装一个指定版本的就行了。
pip uninstall protobuf
pip install protobuf==3.20.0
再次试一试：
import tensorflow as tf
tf.test.is_gpu_available()
可以了：
>>> import tensorflow as tf
2022-11-25 23:32:36.477625: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
>>> tf.test.is_gpu_available()
2022-11-25 23:33:02.299820: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2022-11-25 23:33:02.304514: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library nvcuda.dll
2022-11-25 23:33:02.343180: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties:
name: NVIDIA GeForce GTX 1650 major: 7 minor: 5 memoryClockRate(GHz): 1.755
pciBusID: 0000:01:00.0
2022-11-25 23:33:02.343310: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
2022-11-25 23:33:02.436275: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_100.dll
2022-11-25 23:33:02.553674: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_100.dll
2022-11-25 23:33:02.626239: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_100.dll
2022-11-25 23:33:02.715330: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_100.dll
2022-11-25 23:33:02.785996: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_100.dll
2022-11-25 23:33:03.022703: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2022-11-25 23:33:03.022866: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0
2022-11-25 23:33:03.502224: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1159] Device interconnect StreamExecutor with strength 1 edge matrix:
2022-11-25 23:33:03.502344: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1165] 0
2022-11-25 23:33:03.502842: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1178] 0: N
2022-11-25 23:33:03.503103: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1304] Created TensorFlow device (/device:GPU:0 with 2903 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce GTX 1650, pci bus id: 0000:01:00.0, compute capability: 7.5)
True
>>>
https://blog.csdn.net/qq_19313495/article/details/125049344 "新装的TensorFlow出现protobuf相关错误" ↩︎
