OpenAI gym environment for training agents on Wordle
Project description
gym-wordle
A wordle environment for openai/gym
Installation
Install openai-gym
then install this package with
pip install -e .
Usage
import gym
import gym_wordle
env = gym.make('Wordle-v0')
See the docs for more info
Environment details
This environment simulates a game of wordle using a wordle python clone from https://github.com/bellerb/wordle_solver
The action space is a discrete space of 12972 numbers which corresponds to a word from a list of all allowed wordle guesses and answers
The observation space is a dict with the guesses and colors for the current game.
Guesses is an array of shape(5,6) #5 letters and 6 rows
where each element is a number from 0-26 where 0 is ''
and 26 is z
Colors is an array of the same shape, only each element is a number from 0-2 where 0 is a blank (or black or grey) square, 1 is a yellow square, and 2 is a green square
The reward calculation is as follows:
The agent gets 1-6 points depending on how fast the agent guesses the word. For example, getting the word on the first guesses rewards 6 points, getting the word on the second guess rewards 5 points, etc
The agent also is rewarded for colors in the current row (so the current guess). Right now, the agent is rewarded 3 points for each green tile, and 1 point for each yellow tile. No points are giving for grey tiles
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for alt_gym_wordle-0.7.2-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | ba578a64a832b499705fc5e5ddb50f812510373fe20e8a0e882d9ab3482de462 |
|
MD5 | 2ec37bb05dbcf53462a5be6fd633a9f5 |
|
BLAKE2b-256 | 43a7b7993e7164d2c4986fcb0d6e13770eea3f370c6b1ccb826491faff265731 |