Skip to content

Latest commit

 

History

History
22 lines (20 loc) · 762 Bytes

readme.md

File metadata and controls

22 lines (20 loc) · 762 Bytes
Raw

Algorithms implemented in each folder

  • airl: AIRL
  • apex: Ape-X for distributed training
  • daac: Decoupled advantage actor-critic
  • dqn: Raibow(except c51)
  • dreamer: Dreamer for DMC
  • fqf: FQF
  • gail: GAIL
  • iqn: IQN
  • mrdqn: masked recurrent DQN with retrace(𝝀) and 𝛽-LOO(from Reactor)
  • mriqn: masked recurrent IQN with retrace(𝝀)
  • ppg: PPG with FNN
  • ppo: PPO with FNN
  • ppo2: PPO with masked lstm
  • rnd: Random network distillation
  • sac: SAC, TAC
  • sacd: SAC for discrete action space and CNN
  • sacdiqn: SAC + IQN for discrete action space and CNN
  • seed: SEED for distributed training

Note: Do not run value-based algorithms on the 'procgen' suite. Instead, run the Ape-X version and fill in the corresponding *config.yaml as you need.