|
--- |
|
tags: |
|
- Reinforcement Learning |
|
- PongNoFrameskip-v4 |
|
- deep-reinforcement-learning |
|
model-index: |
|
- name: DQN |
|
results: |
|
- task: |
|
type: reinforcement-learning |
|
name: reinforcement-learning |
|
dataset: |
|
name: PongNoFrameskip-v4 |
|
type: PongNoFrameskip-v4 |
|
metrics: |
|
- type: mean_reward |
|
value: '19' |
|
name: mean_reward |
|
verified: false |
|
--- |
|
|
|
# **DQN** Agent Playing **Pong** |
|
This is a trained model of **DQN** agent that plays **PongNoFrameskip-v4** |
|
Pong is a Atari 2600 game imported from Gym environment. |
|
Agent is implemented from Deep Reinforcement Learning by Max Lapan. |
|
The code is present in the github link: https://github.com/mohit-ix/DeepRL/tree/main/Unit%206 |
|
The performance of agent at different steps is present here: https://youtu.be/03Pl5Odc2jM |
|
|
|
To use the agent use "03_dqn_play.py" from the github link and type: |
|
```python |
|
python 03_dqn_play.py -m [model_name] -r [recording_location] --no-vis |
|
|
|
``` |
|
|
|
Add "-r [recoding_location]" if you want to save the recording and remove "--no-vis" if you want to render the gamplay by the agent. |
