Setup

Setup Instructions

As Saltie includes and uses RLbot it will not work on Mac or Linux (RLBot only runs on Windows)

Everybody

• Install Python 3
• Install the requirements from the requirements.txt

The following commands will be executed in cmd (you might need to run it as administrator)

Windows + R => execute cmd

To use pip in cmd, ensure you the python directroy has been added to your PATH

Instructions for use of the Saltie examples using Pytorch

• Go to Pytorch install locally
• Select your preferred package manager and python version
• Select cuda -> None (the examples do not use the gpu)
• Execute the first suggested command in cmd (the examples do not use torchvision)

Instructions for use of the Saltie examples using Tensorflow

• Install Tensorflow (CPU minimum, GPU recommended):

  • CPU => pip install tensorflow

  • GPU => pip install tensorflow-gpu.

    • If using the GPU version, make sure you have a supported NVIDIA Card and that your CUDA Driver up to date.

  • Ensure you have the required CUDA dlls (added to your PATH directory).

    • If you have Tensorflow = 1.5, use: CuDNN 7; CUDA 9
    • If you have Tensorflow < 1.5, use: CUDNN 6; CUDA 8
    • Link to CUDA Toolkit releases

• Install requests: pip install requests

Streamers

Optional

You can get cool graphs displaying statistics on bot performance (to make up to your viewers for making them endure you). To display these graphs, you will need pyqtgraph and PyQt5.

• Install pyqtgraph: pip install pyqtgraph
• Install PyQt5: pip3 install PyQt5

Run the fancy graph in a separate window by running python live_graph.py from the livedata directory. This can be run before or after you run python runner.py separately.

Trainers

Optional

If you would like to train your own bot you will need to install tqdm for timing info.

• Install tqdm: pip install tqdm

You will also need tensorflow v1.5.

Pull the master repo

Should be self-explanatory. If not, just click here, press the green button over there, download the zip and extract it.

Developers' common issues

Most of us are using an IDE and starting the project at root level. The trainers are not at root level so an IDE is suggested to make it easier to run the trainers. If you run them and get an import error saying 'bot_code' does not exist an alternative option is to add the entire saltie directory to the python path.

If you encounter any other errors, please have a look at the FAQ wiki page, the error solving page, or join the Discord if it didn't help you out!

---

Running Instructions (OUT OF DATE)

Change your username

As you run runner.py, you will record and upload data. We have a leaderboard showing who has created the most replays. To have your name displayed on it, change [User Info] username = unknown in rlbot.cfg to your desired username. This is optional but it will your username instead of unknown on the leaderboard.

Check out your position on the Saltie Stats leaderboard.

From here, you can refer to RLBot for Debugging and Runtime issues

• Start Rocket League.
• Execute RLBot_Injector.exe(in the root folder) as admin. A window should pop up confirming activation.
• Then you can start an exhibition match, optionally with defined parameters.
• Execute python runner.py in your root folder. This runs Saltie with the default parameters.
• Enjoy!