| # stable-baselines3 configuration | |
| # algo: "PPO" | |
| # env_id: "CaribouScipy" | |
| # config: {} | |
| # n_envs: 12 | |
| # tensorboard: "../../../logs" | |
| # total_timesteps: 1000000 | |
| # use_sde: True | |
| # repo: "boettiger-lab/rl4eco" | |
| # save_path: "../saved_agents" | |
| # id: "3" | |
| algo: "PPO" | |
| total_timesteps: 5000000 | |
| algo_config: | |
| tensorboard_log: "../../../logs" | |
| # | |
| policy: 'MlpPolicy' | |
| policy_kwargs: "dict(net_arch=[64, 32])" | |
| # policy_kwargs: "dict(net_arch=[256, 128])" | |
| # batch_size: 512 | |
| # gamma: 0.9999 | |
| # learning_rate: !!float 7.77e-05 | |
| # ent_coef: 0.00429 | |
| # clip_range: 0.1 | |
| # gae_lambda: 0.9 | |
| # max_grad_norm: 5 | |
| # vf_coef: 0.19 | |
| # use_sde: True | |
| # policy_kwargs: "dict(log_std_init=-3.29, ortho_init=False, net_arch=[256, 128])" | |
| # in policy_kwargs: net_arch=[400, 300] | |
| # policy: 'MlpPolicy' | |
| # use_sde: True | |
| # policy_kwargs: "dict(log_std_init=-3, net_arch=[400, 300])" | |
| # clip_range: 0.1 | |
| # env | |
| env_id: "CaribouScipy" | |
| #this is what is passed into the creation of the environment, could pass diff param values here to change the defaults (instead of messing with the params defined in caribou_ode.py) | |
| config: | |
| Tmax: 800 | |
| budget: 50 | |
| n_envs: 12 | |
| tensorboard: "/home/rstudio/logs" | |
| #total_timesteps: 50000 | |
| total_timesteps: 10000000 | |
| use_sde: True | |
| repo: "boettiger-lab/rl4eco" | |
| save_path: "../saved_agents" | |
| id: "LFupdate_shortTF1strat" | |
| #rppo - ppo that has memory (should be better than PPO but hard to train lol) | |
| #td3 - giving some weird errors | |
| #tqc is actually better if we can give it enough time to run | |
| # misc | |
| additional_imports: ["torch"] | |