Anaconda

Python + GPUenvironment for Deep Learning

Make the python + GPU environment necessary to run the Deep Learning program.

[Prerequisite Conditions]

• CUDA (CUDA 9, cuDNN) is installed on Windows.
• Anaconda is installed on Windows.
• Switchable python environment can be created with Conda.

Creating the python + GPU environment

Use tensorflow-gpu instead of tensorflow. Other than that, create the python environment in exactly the same way as "Python Environment for Deep Learning".

1. It is assumed that the environment of python 3.X has been installed on Anaconda. (X ≥ 7)
2. Start "conda", and do the following steps in the "conda" window.
3. Create a new python environment with conda. Here, the environment name is deep. At this moment, tensorflow runs only on python 3.6 or 3.5, so using 3.6.

    (base) C:\Users\nitta>  conda create -n deep python=3.6 jupyter 

4. Switch the new python environment.

    (gpu) C:\Users\nitta>  conda activate deep  

5. Install tensorflow-gpu .

    (gpu) C:\Users\nitta>  pip install --ignore-installed --upgrade tensorflow-gpu  

6. There may be a message saying to update the pip during installation of tensorflow. In that case, update pip according to the instructions.
7. Install keras .

    (gpu) C:\Users\nitta>  pip install keras  

8. Install convenient modules .

    (gpu) C:\Users\nitta>   pip install scipy scikit-learn pillow h5py matplotlib pandas  

9. Install OpenCV .

    (gpu) C:\Users\nitta>  conda install opencv  

10. Install Natural Language ToolKit .

    (gpu) C:\Users\nitta>  pip install nltk  

11. Install gensim for word2vec .

    (gpu) C:\Users\nitta>  pip install gensim