metadata
tags:
- Taxi-v3
- q-learning
- reinforcement-learning
- deep-rl-course
model-index:
- name: q-learning
results:
- task:
type: reinforcement-learning
name: reinforcement-learning
dataset:
name: Taxi-v3
type: Taxi-v3
metrics:
- type: mean_reward
value: 8.23 +/- 2.49
name: mean_reward
verified: false
Q-Learning Agent playing Taxi-v3 🚕
This is a trained Q-Learning agent playing Taxi-v3. This model was trained as part of the Hugging Face Deep RL Course Unit 2.
Training Hyperparameters
n_training_episodes = 25000
learning_rate = 0.7
gamma = 0.95
max_epsilon = 1.0
min_epsilon = 0.05
decay_rate = 0.0005
Evaluation Results
Mean reward: 8.23 +/- 2.49
Usage
import numpy as np
import gymnasium as gym
# Load the Q-table
qtable = np.load("qtable.npy")
# Create environment
env = gym.make('Taxi-v3')
# Run an episode
state, _ = env.reset()
done = False
total_reward = 0
while not done:
action = np.argmax(qtable[state])
state, reward, terminated, truncated, _ = env.step(action)
done = terminated or truncated
total_reward += reward
print(f"Total reward: {total_reward}")