Gradually Building GANs

Gradually building GANs Covering:

• linear-gans without BatchNorm
• linear-gans with batchNorm
• DCGANs with batchNorm
• DCGANs with spectral normalization

Installation (Debian)

• Avoid installing python 3.6.0 It's too old and have lots of error with the new packages.
• If you have py=3.6.15 and a tool for managing virtual environments installed you can skip to step 5 in the installation process.

1. Installing the required python version (python 3.6.15)

• Install dependencies:
  • $ sudo apt update && sudo apt install libssl-dev openssl libbz2-dev liblzma-dev
• Building Python (specifing the folder to build on it)
  • Download the required version:
    • $ wget https://www.python.org/ftp/python/3.6.0/Python-3.6.15.tgz
  • Unzipping the files:
    • $ tar zxf Python-3.6.15.tgz
  • Change dir:
    • $ cd Python-3.6.15/
  • Specify the folder to build on it:
    • create the folder:
      • $ sudo mkdir /opt/python-3.6
    • specifying the folder for installation:
      • $ ./configure --prefix=/opt/python-3.6
  • $ make
  • $ sudo make install Waitting for the build to finish 😅😅😅😅
  • Use this version $ /opt/python-3.6/bin/python3.6 Great you have installed and run the specified py version 😎😎

2. Installing virtualenv: Refer to this gist

3. Creating and activating the venv:

• Create:
  • $ virtualenv --python=/opt/python-3.6/bin/python3.6 ~/venvpy/torch-cpu-py3.6
    • $ --python specifing the directory for the python version $ ~/venvpy/torch-cpu-py3.6 the full path for the venv including the name
• Activate :
  • (fish shell) $ . ~/venvpy/torch-cpu-py3.6/bin/activate.fish
  • (Bash shell) $ source ~/venvpy/torch-cpu-py3.6/bin/activate

4. Upgrade pip:

• $ python -m pip install --upgrade pip make sure that your venv is activated

5. Installing packages:

1. Make sure to run the below command at first for installing the right version of pytorch:
   • $ pip3 install torch==1.10.0+cpu torchvision==0.11.1+cpu torchaudio==0.10.0+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
2. Then run the below command for installing other packages:
   • $ pip install -r requiremnets.txt