[dynamic link library (DLL) initialization example program failed "problem]

Dynamic link library （DLL） Failed to initialize the example program ” problem

---

One 、 Check out the current version anoconda Corresponding available tensorflow

Other links and passes Anaconda The environment is configured in the same way ;
Just installing TensorFlow Do not use directly pip install --upgrade --ignore-installed tensorflow-gpu command ,
But according to anaconda search -t conda tensorflow, Will show the currently available tensorflow edition , Demand selection TensorFlow Version installed .

————————————————

Carry out orders anaconda show conda-forge/tensorflow Get the installation command of the corresponding version conda-forge/tensorflow Select the version name for
Execute the installation command obtained in the above figure conda install --channel http://conda.anaconda.org/conda-forge tensorflow, Installation starts

Two 、 Above unresolved , View the current CPU and cuda Whether the requirements are met

• Build from source on Window Tensorflow Configuration of URL

3、 ... and 、 Above unresolved , Replace source whl file

Tensorflow Installed in the windows It shows ImportError: DLL load failed: Dynamic link library (DLL) Initialization routine failed .

• secsilm_tensorflow After installation import Failure

The reason is that python 3.6.1 32 Bit and 64 The problem of bit ,shell The runtime is still loaded 32 position python Installation package , So we need to find a way to solve it 32 Bit and 64 position shell The problem of simultaneous coexistence , because python3.6.1 32 Bit has written a lot of software , Cannot uninstall .
This problem is solved , The reason is simple , The file used to define the test cannot be named tensorflow.py, Just change to another name .python32, python64 The guaranteed paths on the same computer will not conflict , No additional changes are required .
————————————————
Copyright notice ： This sentence is CSDN Blogger 「fhqlongteng」 The original article of

Four 、 Above unresolved , Download lower version tensorflow1.0（1.10 The following versions ） in the future , Execute the following commands in turn ：

pip install nn
pip install layer
pip install keras
 After the above operations ,tensorflow Fully functional , Can be taken as 1.14 Version USES 

This article takes Code kid blog Understanding learning

5、 ... and 、 There is also a risky solution .

Reinstall the lower version anaconda Adapt it tensorflow1.10 Above version , But the lower version anaconda Are not compatible tensorflow2.0

6、 ... and 、 None of the above has been solved , Please change to a higher configuration CPU