python

For the examples to work properly you need to install ViZDoom system-wide with pip or create a link to dir with ViZDoom pip_package or set doom_game_path and vizdoom_path manually or in config files.

Troubleshooting

• SystemError: dynamic module not initialized properly may mean that you are trying the run an example using python3 instead of python2.

Examples

basic.py

Demonstrates how to use the most basic features of the environment. It configures the engine, and makes the agent perform random actions. It also prints the current state and the reward earned with every action.

buffers.py

cig_multiplayer.py, cig_multiplayer_host.py

Demonstrates how to configure and play multiplayer game for CIG competition.

cig_multiplayer_bots.py

Demonstrates how to play with bots to simulate multiplayer game. Helpful for developing AI agent for CIG competition.

cig_singleplayer.py, cig_multiplayer_host.py

Demonstrates how to configure and play singleplayer game for CIG competition.

delta_buttons.py

Shows how delta buttons work (they may take values other than 0 and 1 and can be used for precise movement).

format.py

Presents different formats of the screen buffer. OpenCV is used to display the images.

fps.py

Tests the performance of the environment in frames per second. It should give you some idea how fast the framework works on your hardware.

labels.py

learning_pytorch.py, learning_pytorch_test.py

Contains an example of how to implement basic Q-learning on the interface within PyTorch

learning_theano.py

Contains an example of how to implement basic Q-learning on the interface within Theano

learning_tensorflow.py

Contains an example of how to implement basic Q-learning on the interface within Tensorflow

multiple_instances.py, multiple_instances_advance.py

Contains an example of how to create a "server" and have multiple agents playing on the server at once. Combine with bots.py and learning_x.py to train agents against some AI

oblige.py (level generator)

Contains an example of how to combine PyOblige with ViZDoom.

record_episodes.py

record_multiplayer.py

scenarios.py

Presents different scenarios that come with ViZDoom environment.

seed.py

Shows how to run deterministic episodes by setting the seed. After setting the seed every episode will look the same (if the agent behaves deterministically).

shaping.py

Demonstrates how to make use of the game variables to implement shaping using health_guided.wad scenario.

spectator.py

Shows how to use the SPECTATOR mode in which YOU play Doom and AI is the spectator (intended for apprenticeship learning).

ticrate.py