README.md

metadata

tags:
  - SuperMarioBros-v0
  - RND
  - CNN
  - reinforcement-learning
  - custom-implementation
  - atari
  - SuperMarioBros-v0
model-index:
  - name: RND-SuperMarioBros-v0
    results:
      - task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: SuperMarioBros-v0
          type: SuperMarioBros-v0
        metrics:
          - type: mean_reward
            value: 2502.70 +/- 418.86
            name: mean_reward
            verified: false

RND with CNN Agent playing SuperMarioBros-v0

This is a trained model of a RND-CNN agent playing SuperMarioBros-v0 . To learn to use this model and train yours check this notebook on kaggle: https://www.kaggle.com/code/syedjarullahhisham/drl-extra-personal-unit-5-rnd-super-mario-bros

Codes

Github repos(Give a star if found useful):

• https://github.com/hishamcse/DRL-Renegades-Game-Bots
• https://github.com/hishamcse/Advanced-DRL-Renegades-Game-Bots
• https://github.com/hishamcse/Robo-Chess

Kaggle Notebook:

• https://www.kaggle.com/code/syedjarullahhisham/drl-extra-personal-unit-5-rnd-super-mario-bros
• https://www.kaggle.com/code/syedjarullahhisham/drl-extra-personal-unit-5-rnd-montezuma-mario-bros

HyperParameters:

"trainmethod": "RND",
"envid": "SuperMarioBros-v0",
"maxstepperepisode": 18000,
"learningrate": 0.0001,
"numenv": 128,
"numstep": 128,
"gamma": 0.999,
"intgamma": 0.99,
"lambda": 0.95,
"usegae": true,
"clipgradnorm": 0.5,
"entropy": 0.001,
"epoch": 4,
"minibatch": 4,
"ppoeps": 0.1,
"extcoef": 5.0,
"intcoef": 1.0,
"stickyaction": true,
"actionprob": 0.25,
"lifedone": false,
"obsnormstep": 50