lsdyna/q-Taxi-v4

Q-Learning agent playing Taxi-v4

This repository contains a tabular Q-Learning agent trained on Taxi-v4.

Evaluation

Episodes	Mean reward	Standard deviation
100	7.5600	2.7067

Files

• q-learning.pkl: Q-table and training hyperparameters.
• results.json: evaluation metrics produced by the training script.
• replay.mp4: one greedy-policy replay generated from the saved Q-table.
• train_taxi.py: training and evaluation source code.

Load the model

import pickle

from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="lsdyna/q-Taxi-v4",
    filename="q-learning.pkl",
)
with open(model_path, "rb") as model_file:
    model = pickle.load(model_file)

qtable = model["qtable"]