Single-file truly minimal implementation of state-of-the-art reinforcement learning algorithms.
Simplicity is the ultimate sophistication. Leonardo da Vinci
| Algorithm | |
|---|---|
| A2C | ❌ |
| ARS | ❌ |
| C51 | ✔️ |
| DDPG | ✔️ |
| Double DQN | ❌ |
| DQN | ✔️ |
| Dueling DQN | ✔️ |
| FQF | ❌ |
| IQN | ✔️ |
| RAINBOW | ❌ |
| Recurrent PPO | ❌ |
| REINFORCE | ✔️ |
| PER (1) | ✔️ |
| PPO | ✔️ |
| QR-DQN | ❌ |
| SAC | ✔️ |
| TD3 | ❌ |
| TQC | ❌ |
| TRPO | ❌ |
(1) with DQN
Most implementations are taken from vwxyzjn/cleanrl without being stricly equivalent. Other ressources: toshikwa/fqf-iqn-qrdqn.pytorch
are welcome.