Akshaykumar4321/rl_course_vizdoom_health_gathering_supreme

Sample-Factory PPO Agent Playing doom_health_gathering_supreme

This is a trained model of a Sample-Factory PPO agent playing doom_health_gathering_supreme.

Evaluation Results

• Environment: doom_health_gathering_supreme
• Mean Reward: 6.85 +/- 1.15
• Target Threshold: >= 5.0 (Certified Pass)

Replay Video

Below is a demonstration of the trained agent navigating the 3D ViZDoom environment to collect health packs:

Training Configuration

{
  "algo": "PPO",
  "env": "doom_health_gathering_supreme",
  "experiment": "doom_health_gathering_supreme",
  "train_for_env_steps": 4000000,
  "num_workers": 8,
  "num_envs_per_worker": 4,
  "batch_size": 2048,
  "gamma": 0.99,
  "gae_lambda": 0.95,
  "learning_rate": 0.0001,
  "encoder_custom": "make_vizdoom_encoder",
  "use_rnn": true,
  "rnn_type": "gru",
  "rnn_num_layers": 1
}