Deep Learning Example Programs

Nguyen Hai Duong
Professor Kim Soo Hyung
Chonnam National University

Environment Setup 1 (with GPUs)

0. Check whether you GPU has Compute Capability 3.0 or higher. If not, please refer either Environment Setup 2 (with GPU Tesla K80 + Google Colab) or Environment Setup 3 (with CPUs)
1. Download and install CUDA® Toolkit 8.0
2. Download cuDNN v5.1, extract downloaded file, copy bin, include, and lib folders to %ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v8.0
3. You should restart your computer to apply system’s changes
4. Download and install Anaconda 4.2.0 with Python 3.5
5. Open Command Prompt as Administrator
6. Install TensorFlow-GPU by entering conda install -c anaconda tensorflow-gpu
7. Enter conda install -c anaconda keras-gpu to install Keras-GPU

Environment Setup 2 (with GPU Tesla K80 + Google Colab)

A free of charge way to experience training deep models with high-performance GPU!

1. Visit Google Colaboratory
2. Sign in with you personal Google account
3. Menu File > New Python 3 notebook
4. Menu Runtime > Change runtime type > Hardware acceleration > GPU > Save
   (Note that python source codes will be saved in your Google Drive, and a work shift lasts for 12 hours)
5. Let's get started with this Jupyter Notebook (how to use GPU, execute Linux commands, install Python packages, and read data from Google Drive)

Environment Setup 3 (with CPUs)

1. Download and install Anaconda 4.2.0 with Python 3.5
2. Open Command Prompt as Administrator
3. Install TensorFlow by entering conda install -c conda-forge tensorflow
4. Enter conda install -c conda-forge keras to install Keras

Usage

1. Download .ipynb files to your computer. For example, D:\dl
2. Open Command Prompt, type cd /d D:\dl
3. Enter jupyter notebook
4. Jupyter IDE will be opened in a web browser, open one of the downloaded programs
5. Select menu Cell > Run All to run the program