Skip to content

Latest commit

 

History

History
122 lines (78 loc) · 5.36 KB

Install.md

File metadata and controls

122 lines (78 loc) · 5.36 KB
Raw

Prerequisites

  • Windows / Linux
  • Python >= 3.7
  • A recent NVIDIA GPU (required only on the training computer if you plan to train your own models)

If using Anaconda on Windows:

We recommend installing the conda version of pywin32:

conda install pywin32

Installation

(Note for ML developers: in case you are not interested in using support for TrackMania, you can simply install the tmrl library).

The following instructions are for installing tmrl with support for the TrackMania 2020 video game.

You will first need to install TrackMania 2020 (obviously), and also a small community-supported utility called Openplanet for TrackMania (the Gymnasium environment needs this utility to compute the reward).

Windows users - Install TrackMania 2020:

(Required only on the computer(s) running TrackMania)

To install the free version of TM20, you can follow the instructions on their official website .

Windows users - Install Openplanet:

(Required only on the computer(s) running TrackMania)

Make sure you have the Visual C++ runtime installed or OpenPlanet will not work. You can download it here for 64bits versions of Windows.

Then, install Openplanet for TrackMania.

During the installation, Windows may complain that OpenPlanet has no valid certificate (this is a small non-commercial tool not signed by any company). In such case, you will have to hit the link for "more info", and then click "install anyway".

Linux users:

(Windows users can skip this section)

Since version 0.6.0, we support the full TrackMania 2020 pipeline on Linux, including the gymnasium environment. Because Ubisoft Nadeo does not officially support Linux, we wrote a Linux tutorial to help you set up TrackMania and OpenPlanet on your machine.

Install TMRL:

To install the tmrl python library, open your favorite terminal and run:

pip install tmrl

Then, validate the installation:

python -m tmrl --install

Additional information for Windows / Trackmania 2020:

If running on Windows, during the installation, a driver will be installed to emulate a virtual gamepad. Accept the licence agreement and install the driver when prompted.

Image

Then, navigate to your home folder (on Windows it is C:\Users\your username\).

There, you will find that tmrl has created a folder named TmrlData.

On the computer(s) running TrackMania, OpenPlanet should also have created a folder named OpenplanetNext there. (If OpenplanetNext is not there, launch Trackmania after installing Openplanet, and it should be created automatically).

Open the OpenplanetNext\Plugins folder and double-check that pip has copied TMRL_GrabData.op there. If not, navigate to TmrlData\resources, copy the Plugins folder, and paste it in the OpenplanetNext folder.

(NB: when pip-uninstalling tmrl, the TmrlData folder is not deleted.)

Clean install:

If at some point you want to do a clean re-install of tmrl:

  • pip uninstall tmrl
  • Delete the TmrlData folder from your home folder
  • pip install tmrl

Set up TMRL

(Optional) Configure/manage TMRL:

The TmrlData folder is your "control pannel", it contains everything tmrl uses and generates:

  • The checkpoints subfolder is used by the trainer process: it contains persistent checkpoints of your training,
  • The weights subfolder is used by the worker process: it contains snapshots of your trained policies,
  • The reward subfolder is used by the worker process: it contains your reward function,
  • The dataset subfolder is for RL developers (to use with custom replay buffers),
  • The config subfolder contains a configuration file that you probably want to tweak.

Navigate to TmrlData\config and open config.json in a text editor.

( ℹ️ config.json is described in details here.)

In particular, you may want to adapt the following entries:

  • RUN_NAME: set a new name for starting training from scratch
  • LOCALHOST_WORKER: set to false for workers not on the same computer as the server
  • LOCALHOST_TRAINER: set to false for trainer not on the same computer as the server
  • PUBLIC_IP_SERVER: public IP of the server if not running on localhost
  • PORT needs to be forwarded on the server if not running on localhost
  • WANDB_PROJECT, WANDB_ENTITY and WANDB_KEY can be replaced by you own wandb credentials for monitoring training

You can delete the content of all folders (but not the folders themselves) whenever you like (except config.json, a default version is provided in resources if you delete this).

To reset the library, delete the entire TmrlData folder and run:

python -m tmrl --install

This will download and extract the TmrlData folder back to its original state.

(Optional) Check that everything works:

Launch TrackMania 2020, launch a track, then press f3 to open the OpenPlanet menu, open the logs by clicking OpenPlanet > Log, and in the OpenPlanet menu click Developer > (Re)load plugin > TMRL Grab Data. You should see a message like "waiting for incoming connection" appear in the logs. Press f3 again to close the menu.