Get Started
In this tutorial, we will install the tools, set up the build environment, create your first PyTorch Live project and see it running in an emulator or your phone.
PyTorch Live's command line tool torchlive-cli
will automate most of the setup process. Here's a sneak peek of the PyTorch Live app when the setup is complete:
Let's get started!
Prerequisites
- Our beta release only supports MacOS. Windows and Linux supports are coming soon. However, you can follow the manual installation instructions to get started on Windows and Linux.
- Node.js: If you don't have it already, you can download Node.js LTS from its website or install via Homebrew (ie,
brew install node
).- Note that Node.js version 17 can cause issues, so it is recommended to use any of the LTS versions 12, 14, or 16.
- Xcode: If you want to run the app on an iOS simulator or a physical iOS device, you need to install Xcode from the Apple App Store on your macOS device. PyTorch Live currently supports Xcode version 12.5 or later.
Installation
Setting up the build environment can be tedious and difficult. PyTorch Live provides a setup routine to install the required dependencies automatically for you. It will try to install the following libraries and tools:
- Homebrew
- OpenJDK
- Yarn
- Watchman
- Android Command Line Tools
- Android SDK
- Android SDK Manager
- Android Virtual Device Manager
- Android Emulator
- CocoaPods
The PyTorch Live CLI currently does not support the Apple M1 chipset. However, there are instructions by the community for how to install React Native on an Apple M1 without Rosetta.
Here's what you will see during the automatic setup process:
Check out the manual installation process if your operating system is Windows or Linux or if the PyTorch Live command line tool torchlive-cli
fails for you!
Now let's run this setup-dev command and get the build environment ready for development:
❯ npx torchlive-cli setup-dev
You will need administrator access (using sudo
) to run this command. It may take 20 minutes to complete, depending on your laptop configurations.
You only need to run this once to install the dependencies. No need to run this again to start a new project.
Check installations
If everything goes well, the terminal will looks similar to the following output. If you run into any issues, please check the FAQ.
❯ npx torchlive-cli setup-dev
_ _ _ _
| |_ ___ _ __ ___| |__ | (_)_ _____
| __/ _ \| '__/ __| '_ \| | \ \ / / _ \
| || (_) | | | (__| | | | | |\ V / __/
\__\___/|_| \___|_| |_|_|_| \_/ \___|
torchlive version 0.0.2-alpha.5
Password:
↓ Homebrew (3.1.7) [SKIPPED]
✔ OpenJDK (1.8.0)
✔ Watchman (4.9.0)
↓ Node (16.1.0) [SKIPPED]
✔ Yarn (1.22.10)
✔ Android SDK
✔ Android SDK Manager
✔ Android Emulator
✔ Android Emulator Skin
Check Python version
Currently, python3
is required to bootstrap PyTorch Mobile models used in the example app. In macOS Catalina and Big Sur, python3
is pre-installed. Run the following command to check if python3 is installed:
❯ python3 --version
Python 3.9.4
The version installed on your machine might be slightly different, but should at least be version 3.7
or above.
If python3
is not installed, you can install it via Homebrew on macOS:
brew install python3
Create a New Project
Now that we have set up the build environment, let's create a new PyTorch Live project and see it running in an emulator or on your device. It's as simple as running two commands using the torchlive-cli
tool.
Initialize Your Project
For PyTorch Live beta, we provide a bootstrap React Native app, packed with AI demos and practical examples. Use the npx torchlive-cli init
command to initialize your first PyTorch Live project.
npx torchlive-cli init MyFirstProject
Behind the scene, the tool created a new React Native project and configured the build tools for mobile development. It also added a couple on-device AI models as well as examples to run them.
This is what you should see after the project is initialized:
❯ npx torchlive-cli init MyFirstProject
_ _ _ _
| |_ ___ _ __ ___| |__ | (_)_ _____
| __/ _ \| '__/ __| '_ \| | \ \ / / _ \
| || (_) | | | (__| | | | | |\ V / __/
\__\___/|_| \___|_| |_|_|_| \_/ \___|
torchlive version 0.0.2-alpha.5
✔ project MyFirstProject
Run Your Project
We are almost there. Now let's navigate to the newly created PyTorch Live project folder:
cd MyFirstProject
And run the PyTorch Live project on Android:
npx torchlive-cli run-android
By default, the PyTorch Live app will run in an emulator. You can also run it on your own phone which may improve performance. To see it on your phone, plug it into your computer and enable Android's developer mode.
On Android 4.1 and lower, the Developer options screen is available by default. On Android 4.2 and higher, you must enable this screen. To enable developer options, go to Settings > Developer options and tap the Build Number option 7 times.
More details are on the Android configure on-device developer options.
Welcome to PyTorch Live
Start by trying out the various Examples in the app. When you are ready to experiment, take a look at the Toolbox demos and follow along with a tutorial: recognizing objects in photographs or answering questions about some text.
PyTorch Live is in beta. Your feedback will help us improve tremendously. Please file issues and share ideas on GitHub.
Show us what you've made with PyTorch Live by tagging #playtorch in your social media posts too.