Git, Python & Conda Guide

• Git
  • Overview
  • Installation
    • Prerequisites
    • PC
      • Command Line Version
      • GUI Version
    • Mac
      • Command Line Version
        • Option A
        • Option B
      • GUI Version
  • Use
    • Creating a Repo
    • Cloning a Repo
      • GUI
      • Command Line
    • Editing a Repo
      • GUI
      • Command Line
• Python
  • Overview
  • Installation
    • PC
    • Mac
• PIP
  • Overview
  • Installation
  • Use
• CONDA
  • Overview
  • Installation
    • PC
    • Mac
  • Use
• Realtime Voice Cloning

Git

Overview

Git is a version control system. It allows you to store code online and to move back and forward between older an newer versions of your code.

Version control with Git goes something like this:

• A Git Repository is the name for a single project. Repo is the short name.
• You can create a new Git Repo locally on your computer or using your Github/Gitlab account.
• You can download a Repo as a zipped folder directly, or you can clone it to your computer using either the terminal/command line or a GUI application.
• Changes made to the downloaded (local) version of the Repo will not affect the version stored online until you commit and push your changes.
• Changes can be pushed to the Master branch of your Repo or to a new branch.
• You can switch easily between your Master branch and other created branches.
• Any changes you push can usually be reversed by rolling back to a previous commit.
• When working with a repo created by another Git user, you can create a fork of the repo, essentially a copy of the repo on your own Git account that you can edit.
• When multiple people are working on a single repo, changes created on new branches can be merged into the master branch.
• This might not all be clear now, you will likely not need to use all of these functions! See Use section below for instructions on how to clone a repository, and how to add, commit and push changes.

Installation

Prerequisites

Create an account on GitHub

PC

Command Line Version

1. Download and install.
   • This is what I tend to use. It comes with a dedicated command line application called Git Bash to run Git commands (e.g. commit, add, clone, etc).
   • NOTE: Even if you plan to use the GUI version below, I recommend downloading these tools. I have found that the GIT Bash terminal useful for Python and Conda (see below).
2. After installation, open the GIT Bash terminal, copy and paste the below code and hit enter (replace 'username' and 'email' with the username and email you used when signing up to GitHub)

   git config --global user.name "username"
   git config --global user.email "email@example.com"
   

GUI Version

1. Download and install.
   • This is a desktop app created by GitHub. It will allow you to do all the same things you can do in the command line version but with a graphical interface that is useful for more easily seeing the changes made to your repo.
2. After downloading, sign in using your GitHub account.

Mac

Command Line Version

Option A

1. Git may already be installed on your Mac. If not, you can install it using the XCode Command Line Tools. To check if it is installed, open the Terminal application, copy and paste the following, and hit enter:

   git --version
   

2. If Git is already installed, you will see something like the image below:

3. If Git is not installed, you should be prompted to install it via XCode. Follow the installation instructions on screen.

4. Add your credentials by opening Terminal, copying and pasting the following, and hitting enter (replace 'username' and 'email' with the username and email you used when signing up to GitHub)

   git config --global user.name "username"
   git config --global user.email "email@example.com"
   

Option B

1. Install Homebrew using the following steps:

   • Open the Terminal application on your Mac
   • Copy and paste the following into the terminal window, then press enter:

   /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install.sh)"
   

2. Install Git using Homebrew

   • When Homebrew has finished installing, copy and paste the following into the terminal window, then press enter:

   brew install git
   

3. Add your credentials by opening Terminal, copying and pasting the following, and hitting enter (replace 'username' and 'email' with the username and email you used when signing up to GitHub)

   git config --global user.name "username"
   git config --global user.email "email@example.com"
   

GUI Version

1. Download and install.
2. After downloading, sign in using your GitHub account.

Use

Creating a Repo

1. Go to Github
2. In your GitHub homepage, click New as shown in the image below:

3. In the next page, set the name of your repo, and a description.

4. You can also add a .gitignore file and a README here. The .gitignore file tells Git which files to ignore when uploading changes. There are a number of .gitignore files available for popular programming languages and libraries, or you can create your own custom .gitignore. See below, you can select an Unreal Engine .gitignore file - this will ignore, for example, files related to caching and backups in the Intermediate and Saved folders.

5. That's it! Repo created. If you want, you can upload files to the next page as shown in the image below (pretty self-explanatory), but in the next section I'll show you how to clone a repo to your computer and add files this way.

Cloning a Repo

GUI

1. Open GitHub Desktop
2. If you have already connected it to your account, you should see an option to Clone a repository from the internet. Click this and you should see a list of repositories already in you account (if you have any) and the URL tab which will allow you to clone repos from other Git users.

Select a repo, or enter a repo URL, set the 'Local Path' (i.e. where you want it to be downloaded to on your computer) and hit clone. The files will be downloaded and you're ready to go.

Command Line

1. Find the URL for the repo you want to clone by visiting the repo page and clicking the green Code button. Copy the https link. (For example, the link for this repo is https://github.com/jolmcn/guide.git )

2. Open Terminal (on Mac) or Git Bash (on Windows).

3. Navigate to the location where you want to download the repo using the cd command (cd = change directory). For example, if you want to download the repo to your Documents use the following to navigate to documents:

cd Documents

4. From Documents, or wherever else you'd like to store your repo (I keep mine in a folder called Developer) clone your repo using git clone followed by the URL of the repo then hit enter. e.g.:

git clone https://github.com/jolmcn/guide.git

5. That's it! Your repo is cloned. The files are downloaded and you're ready to go.

Editing a Repo

If you want changes made locally to your repo to be pushed to the online version, follow these steps:

GUI

1. Open Github Desktop and select your repo. Under the Uncommitted Changes tab you will see the changes you have made to the files in the repo since the last commit. Red lines indicate code that has been deleted, green indicates code that has been added/modified.

2. To commit your changes, add a summary and description of the changes you've made, then click Commit to Master

3. To push your changes to the online repo, click 'Sync' on the top right corner.

Command Line

1. Open Terminal (on Mac) or Git Bash (on Windows).

2. Navigate to your git repo using the cd command

3. Use the following command to add all files to your commit

git add .

4. Use the following command to commit the changes, with a message describing the changes

git commit -m "enter the message here"

5. Use the following command to push the commit to your GitHub account

git push

Python

Overview

Python is a popular, high-level coding language used widely in Machine Learning and AI applications, to name just a few contexts. Most ML projets will require you to run a Python program and/or download some other Python libraries. There are a number of different ways to download existing Python libraries.

Installation

PC

1. Download and run the executable file from the [Python website](Windows x86-64 executable installer) to install Python on your computer.

   • **NB: Make sure that you check the box for 'Add Python 3.7 to PATH' on the first window when you start to install - see image below
   • Note: The above link is for Python 3.7. The newest version is 3.8, but I have had trouble running Tensorflow and PyTorch projects using Python 3.8, so I recommend downloading 3.7

   

2. Open Git Bash and run the following command to check if Python has been successfully installed:

python --version

If it returns something like 'The command python is not recognised' then Python is not installed successfully.

Mac

The most recent version of Python 2 should already be installed on your Mac by default. Open Terminal and run the following command to ensure Python is installed:

python --version

If it returns something like 'the command python is not recognised', then Python is not installed.

To install Python:

1. Install Homebrew (if you have not already installed it as described above) by running the following command in Terminal:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install.sh)"

2. Install Python using Homebrew using the following command in Terminal;

brew install python@3.7

3. Run python --version again to confirm that Python has been installed (if it does not work, try restarting Terminal and running the command again).

PIP

Overview

PIP is the standard package management system for Python. In other words, it is the default tool for installing existing Python libraries. Unless you have created a local environment (e.g. using Miniconda - see below), libraries/packages installed with PIP will be installed globally on your computer.

Installation

PIP should be installed automatically along with Python. Confirm that PIP is installed by running the following command in Terminal (on Mac) or Git Bash (on Windows):

pip --version

Use

To install a Python Package using PIP run the following command in Terminal (on Mac) or Git Bash (on Windows):

pip install package-name

To install a specific version of a Python Package using PIP run the following command:

pip install package-name==1.0.0

To search for a specific Python Package using PIP run the following command:

pip search package-name

To list the Python Packages that have been installed in your current environment run the following command:

pip list

CONDA

Overview

Anaconda/Miniconda is a platform for creating enclosed environments for running Python projects.

Conda allows us to run multiple projects on the same computer with different versions of Python or different versions of specific Python packages. Many existing repos on Github will require different versions of Python or specific Python packages, and Conda can allow us to run these projects more easily on the same device. When you enter a Conda environment, any packages installed via Conda or PIP will only exist in that environment. When that environment is deactivated, the installed packages will not be available.

Installation

PC

1. Download and run the Miniconda Installation File
2. Open the Git Bash application and run one of the following commands to test if Conda has installed correctly (you may need to restart your computer if it returns an error):

conda --version

conda --info

Mac

1. Download and run the Miniconda Installation File
2. Open Terminal and run one of the following commands to test if Conda has installed correctly (you may need to restart your computer if it returns an error):

conda --version

conda --info

Use

See the Conda Cheat Sheet for a full list of commands

All of these commands are to be run in Terminal (on Mac) or Git Bash (on PC).

• Create a new Conda environment (Note: you can name your environment whatever you like):

conda create -n environmentName

• List the Conda environments that you have created:

conda info -e

• Activate a Conda environment after creating it:

conda activate environmentName

• Deactivate the currently activated Conda environment:

conda deactivate

The following commands should only be run when you are in an active Conda environment:

• Install Conda packages using the following command:

conda install package-name

• Install a specific version of a Conda package:

conda install package-name==1.0

• Search for a specific Conda package:

conda search package-name

• List the Conda packages installed in the current active Conda environment:

conda list

Realtime Voice Cloning

Repo Link

1. Open Git Bash and clone the repo to your computer's Documents folder using the following commands in this order:

cd Documents

git clone https://github.com/CorentinJ/Real-Time-Voice-Cloning

cd Real-Time-Voice-Cloning

2. If you do not aready have it installed, download Cuda 10.1 from here using the settings shown in the image below:

3. Install Python virtualenv using pip with the following command in Git Bash:

pip install virtualenv

4. Create a virtualenv inside the repo folder using the following command in Git Bash:

virtualenv env

This creates an environment where any new packages installed will only be used when the environment is activated, similar to Conda above.

6. Active the environment using the following command in Git Bash

. env/Scripts/activate

If you've come this far successfully and the environment is successfully activated, any new lines in Git Bash will be preceded by the environment name (env), and Git Bash should look something like the image below:

4. Install PyTorch by running following command in Git Bash:

pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html

If the above command doesn't work, it may be because you have a different version of Cuda installed. In this case, visit this page and enter the settings applicable to your machine.

5. Install the other packages required for the repo by running the following command in Git Bash

pip install requirements.txt

This should install all of the packages in the requirements.txt file. I had trouble with this, so i had to install them individually! If you also have trouble, try running the below command instead:

pip install tensorflow==1.15 umap-learn visdom librosa matplotlib numpy scipy tqdm sounddevice SoundFile Unidecode inflect PyQt5 multiprocess numba==0.48

6. Download the pretrained models. The unzipped folder will contain three subfolders - encoder, synthesizer, and vocoder. Copy the contents of these folders into the folders with the same names in the repo.

7. Optional Download the LibriSpeech dataset from here. Create a new folder inside the Real-Time-Voice_cloning repo folder, then extract the contents of the downloaded dataset to Real-Time-Voice-Cloning/Datasets/LibriSpeech/train-clean-100 NOTE: If you do not download this dataset you can use your own audio files or record directly into the GUI to generate audio.

8. Finally launch the Toolbox. If you have downloaded the dataset run the following command:

python demo_toolbox.py -d Datasets

If you have not downloaded the dataset, run the following command:

python demo_toolbox.py

You should see the following window open:

• Click on Browse to upload an audio file from your computer of someone speaking, to replicate this voice. If you have downloaded the dataset, you can click Load to load an audio file from the dataset.
• Enter the text you would like to have synthesized in the textbox and click Synthesize and vocode.
• When synthesis and vocoding is complete, you will hear the generated audio play.
• To save the audio, click Export