AndyYue1893 / RL-Implementation

simple code to reinforcement learning

RL-Implementation

simple code to reinforcement learning

• pendulum

  • python -um algos.ppo.run_ppo_torch --exp_name xxx --max_grad_norm 0.5 --steps 2048 --anneal_lr --is_clip_v --seed 20 --env Pendulum-v0 --is_gae --target_kl 0.03 --lr 0.0003 --norm_state --iteration 1000 --batch 64 --last_v --norm_rewards returns --a_update 10

• mujoco

  • python -um algos.ppo.run_ppo_torch --exp_name xxx --max_grad_norm 0.5 --steps 2048 --anneal_lr --is_clip_v --seed 30 --env HalfCheetah-v2 --is_gae --target_kl 0.07 --lr 0.0003 --norm_state --iteration 2000 --batch 64 --last_v --norm_rewards returns --a_update 10

• dmc

  • python -um algos.ppo.run_ppo_dmc_torch --exp_name xxx --max_grad_norm 0.5 --steps 2048 --anneal_lr --is_clip_v --seed 10 --domain_name cheetah --task_name run --network mlp --target_kl 0.05 --lr 0.0003 --iteration 2000 --batch 64 --encoder_type state --last_v --is_gae --a_update 10 --norm_state --norm_rewards returns --c_en 0

• if you want to test dmc and see video

  • python -um algos.ppo.test_ppo_torch --exp_name xxx --encoder_type state

• if you want to plot results (mujoco)

  • please modify taskname in plot_seed.py
  • then python -um algos.ppo.plot_seed

• output results in data/* when you train experiments

  • you can find logs(tensorboard),checkpoints(pytorch),args.json and process.txt

• if you know chinese, you can read

  • 强化学习中的调参经验与编程技巧(on policy 篇) https://blog.csdn.net/qq_27008079/article/details/108313137