Tetris using Learning from Demonstrations

This porject presents the classic game of Tetris built using PyGame whcich makes use of anm agente to learn trough human data how to play this mytical game.

Usage

main.py for playing Tetris manually.

main_imitation.py for training and data collection Imitation learning.

The different modes of use can be applied by modifying the following functions:

if __name__ == "__main__":
    train() # Trains and validate the model using data acquired playing
    model.save_weights() # Save the model weigths
    model.load_weights('_10k_01_nat1') # Load the model weigths
    main() # Shows a model trained or allow the user to play and generate new data
    generate_data(400) #Generates sinthetic data using a natural selection algorithm

If the process you wanna do is take data, you must select in main_imitation.py.

def main(manual=1): # Manual=1(user data acquisition)

Environment

To edit Tetris behaviour change contructor params.

env = Tetris({
  'reduced_shapes': True,
  'reduced_grid': True
})

Where reduced_shapes control the type of pieces used in the learning process and reduced_grid control de type of grid.

Enviorment follows regular OpenAI standard.

from enviorment.tetris import Tetris

from Imitation.agent import imitation_agent
from nat_selection.model import Model
from dqn.agent import DQN

env = Tetris() 

# Agent can either be custom.
 agent = imitation_agent(env)


total_score = 0
state, reward, done, info = env.reset()

while not done:

  # for random action
  action = env.action_sample 
  
  # for agent action
  # Be aware that different agents have different methods of getting the next action.
  action = agent.policy(state)
  
  state, reward, done, info = env.step(action)
  total_score += reward
  
  env.render()

State is a three dimensional array with 2 layers representing the placed blocks and the ones you are controlling. Blocks are 1 and blank cells are 0.

Loading pretrained models can be done like this.

from enviorment.tetris import Tetris

from Imitation.agent import imitation_agent
from nat_selection.model import Model
from dqn.agent import DQN

env = Tetris() 

# Imitation
agent = imitation_agent(env)
agent.load_weights('_10k_01_nat1')
# Where the argument is the suffix after "weights" in ./Imitation/weights

Install

For installation, enter the directory

cd TetrisLfd/

and install al the dependencies:

pip install -r requirements.txt

and Pytorch