metadata
tags:
- CliffWalking-v0
- q-learning
- reinforcement-learning
- custom-implementation
model-index:
- name: qlearning
results:
- task:
type: reinforcement-learning
name: reinforcement-learning
dataset:
name: CliffWalking-v0
type: CliffWalking-v0
metrics:
- type: mean_reward
value: '-13.00 +/- 0.00'
name: mean_reward
verified: false
Q-Learning Agent playing CliffWalking-v0
This is a trained model of a Q-Learning agent playing CliffWalking-v0. The agent was trained for 100000 episodes.
Evaluation Results
- Mean Reward: -13.00 +/- 0.00
Usage
import gymnasium as gym
import pickle
from huggingface_hub import hf_hub_download
def load_from_hub(repo_id, filename):
pickle_model = hf_hub_download(repo_id=repo_id, filename=filename)
with open(pickle_model, 'rb') as f:
downloaded_model_file = pickle.load(f)
return downloaded_model_file
model_data = load_from_hub(repo_id="dllmpg/qlearning", filename="q-learning.pkl")
q_table = model_data["qtable"]
env_id = model_data["env_id"]
# Example of running the loaded agent
env = gym.make(env_id)
raw_state, info = env.reset()
state_idx = raw_state # CliffWalking uses direct state indexing
# ... run agent using greedy_policy(q_table, state_idx) ...