Downloads · 30 days
0
q-learning-Taxi-v3
muhrivandysetiawan/q-learning-Taxi-v3
q-learning-Taxi-v3 is a reinforcement learning model from muhrivandysetiawan. Use it for the reinforcement learning task on the model card, and read the license before you ship it in a product.
Trained with pure NumPy Q-Learning (no deep learning library). Hyperparameters optimized via Optuna TPE -- 20 trials. Project 2/12 of a Deep RL portfolio.
Access
Public
Updated Apr 6, 2026
Repo size
48.3 KB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.mp4196 KB · 79%
At a glance
- Task
- Reinforcement Learning
- Access
- Public
- Created
- Apr 5, 2026
- Updated
- Apr 6, 2026
- SHA
- 30eda63d
Model card
From the Hugging Face model README
Q-Learning Agent -- Taxi-v3
Trained with pure NumPy Q-Learning (no deep learning library). Hyperparameters optimized via Optuna TPE -- 20 trials. Project 2/12 of a Deep RL portfolio.
Performance
| Version | Score | Notes |
|---|---|---|
| Baseline | 5.4865 | Default params |
| Optuna v1 | 5.7480 | 20-trial sweep |
| Target | >= 4.5 | Certification |
| Status | PASS |
Algorithm
Q-Learning from scratch -- pure NumPy, no SB3.
Bellman equation implemented in QTable.update().
Q(s,a) <- Q(s,a) + alpha * [r + gamma * max Q(s',a') - Q(s,a)]
Architecture (OOP)
QTable -- Bellman update, save/load EpsilonGreedyPolicy -- explore/exploit with decay QLearningEnv -- env wrapping, seed control QLearningTrainer -- training loop QLearningEvaluator -- greedy eval, separated from training
Best Hyperparameters (Optuna)
| Param | Value |
|---|---|
| learning_rate (alpha) | 0.4904 |
| gamma | 0.9683 |
| decay_rate | 0.00270 |
| min_epsilon | 0.0028 |
| n_episodes | 70,000 |
Reproduce
from huggingface_hub import hf_hub_download
import numpy as np
qtable = np.load(
hf_hub_download("muhrivandysetiawan/q-learning-Taxi-v3", "qtable_taxi_optuna.npy")
)
# qtable shape: (500, 6)
# best_action = np.argmax(qtable[state])
Portfolio
Project 2/12 -- HuggingFace Deep RL Curriculum Author: muhrivandysetiawan W&B Dashboard: https://wandb.ai/muhrivandysetiawan-muh-rivandy-setiawan/deep-rl-portfolio
- Task
- Reinforcement Learning
- Created
- Apr 5, 2026
- Updated
- Apr 6, 2026