Add --gym-packages parameter to push_to_hub

hyperparams/a2c.yml

@@ -19,6 +19,12 @@ CartPole-v1:
   policy: 'MlpPolicy'
   ent_coef: 0.0
 
+seals/CartPole-v0:
+  n_envs: 8
+  n_timesteps: !!float 5e5
+  policy: 'MlpPolicy'
+  ent_coef: 0.0
+
 LunarLander-v2:
   n_envs: 8
   n_timesteps: !!float 2e5
@@ -35,6 +41,13 @@ MountainCar-v0:
   policy: 'MlpPolicy'
   ent_coef: .0
 
+seals/MountainCar-v0:
+  normalize: true
+  n_envs: 16
+  n_timesteps: !!float 1e6
+  policy: 'MlpPolicy'
+  ent_coef: .0
+
 Acrobot-v1:
   normalize: true
   n_envs: 16
@@ -170,19 +183,39 @@ HalfCheetah-v3: &mujoco-defaults
   n_timesteps: !!float 1e6
   policy: 'MlpPolicy'
 
+seals/HalfCheetah-v0:
+  <<: *mujoco-defaults
+
 Ant-v3:
   <<: *mujoco-defaults
 
+seals/Ant-v0:
+  <<: *mujoco-defaults
+
 Hopper-v3:
   <<: *mujoco-defaults
 
+seals/Hopper-v0:
+  <<: *mujoco-defaults
+
 Walker2d-v3:
   <<: *mujoco-defaults
 
+seals/Walker2d-v0:
+  <<: *mujoco-defaults
+
 Humanoid-v3:
   <<: *mujoco-defaults
   n_timesteps: !!float 2e6
 
+seals/Humanoid-v0:
+  <<: *mujoco-defaults
+  n_timesteps: !!float 2e6
+
 Swimmer-v3:
   <<: *mujoco-defaults
   gamma: 0.9999
+
+seals/Swimmer-v0:
+  <<: *mujoco-defaults
+  gamma: 0.9999

hyperparams/ars.yml

@@ -5,6 +5,12 @@ CartPole-v1:
   policy: 'LinearPolicy'
   n_delta: 2
 
+seals/CartPole-v0:
+  n_envs: 1
+  n_timesteps: !!float 5e4
+  policy: 'LinearPolicy'
+  n_delta: 2
+
 # Tuned
 Pendulum-v1: &pendulum-params
   n_envs: 1
@@ -41,6 +47,11 @@ MountainCar-v0:
   n_delta: 8
   n_timesteps: !!float 5e5
 
+seals/MountainCar-v0:
+  <<: *pendulum-params
+  n_delta: 8
+  n_timesteps: !!float 5e5
+
 # Tuned
 MountainCarContinuous-v0:
   <<: *pendulum-params
@@ -119,6 +130,17 @@ Swimmer-v3:
   alive_bonus_offset: 0
   # normalize: "dict(norm_obs=True, norm_reward=False)"
 
+seals/Swimmer-v0:
+  n_envs: 1
+  policy: 'LinearPolicy'
+  n_timesteps: !!float 2e6
+  learning_rate: !!float 0.02
+  delta_std: !!float 0.01
+  n_delta: 1
+  n_top: 1
+  alive_bonus_offset: 0
+  # normalize: "dict(norm_obs=True, norm_reward=False)"
+
 Hopper-v3:
   n_envs: 1
   policy: 'LinearPolicy'
@@ -130,6 +152,17 @@ Hopper-v3:
   alive_bonus_offset: -1
   normalize: "dict(norm_obs=True, norm_reward=False)"
 
+seals/Hopper-v0:
+  n_envs: 1
+  policy: 'LinearPolicy'
+  n_timesteps: !!float 7e6
+  learning_rate: !!float 0.01
+  delta_std: !!float 0.025
+  n_delta: 8
+  n_top: 4
+  alive_bonus_offset: -1
+  normalize: "dict(norm_obs=True, norm_reward=False)"
+
 HalfCheetah-v3:
   n_envs: 1
   policy: 'LinearPolicy'
@@ -141,6 +174,17 @@ HalfCheetah-v3:
   alive_bonus_offset: 0
   normalize: "dict(norm_obs=True, norm_reward=False)"
 
+seals/HalfCheetah-v0:
+  n_envs: 1
+  policy: 'LinearPolicy'
+  n_timesteps: !!float 1.25e7
+  learning_rate: !!float 0.02
+  delta_std: !!float 0.03
+  n_delta: 32
+  n_top: 4
+  alive_bonus_offset: 0
+  normalize: "dict(norm_obs=True, norm_reward=False)"
+
 Walker2d-v3:
   n_envs: 1
   policy: 'LinearPolicy'
@@ -152,6 +196,17 @@ Walker2d-v3:
   alive_bonus_offset: -1
   normalize: "dict(norm_obs=True, norm_reward=False)"
 
+seals/Walker2d-v0:
+  n_envs: 1
+  policy: 'LinearPolicy'
+  n_timesteps: !!float 7.5e7
+  learning_rate: !!float 0.03
+  delta_std: !!float 0.025
+  n_delta: 40
+  n_top: 30
+  alive_bonus_offset: -1
+  normalize: "dict(norm_obs=True, norm_reward=False)"
+
 Ant-v3:
   n_envs: 1
   policy: 'LinearPolicy'
@@ -163,6 +218,17 @@ Ant-v3:
   alive_bonus_offset: -1
   normalize: "dict(norm_obs=True, norm_reward=False)"
 
+seals/Ant-v0:
+  n_envs: 1
+  policy: 'LinearPolicy'
+  n_timesteps: !!float 7.5e7
+  learning_rate: !!float 0.015
+  delta_std: !!float 0.025
+  n_delta: 60
+  n_top: 20
+  alive_bonus_offset: -1
+  normalize: "dict(norm_obs=True, norm_reward=False)"
+
 
 Humanoid-v3:
   n_envs: 1
@@ -175,6 +241,17 @@ Humanoid-v3:
   alive_bonus_offset: -5
   normalize: "dict(norm_obs=True, norm_reward=False)"
 
+seals/Humanoid-v0:
+  n_envs: 1
+  policy: 'LinearPolicy'
+  n_timesteps: !!float 2.5e8
+  learning_rate: 0.02
+  delta_std: 0.0075
+  n_delta: 256
+  n_top: 256
+  alive_bonus_offset: -5
+  normalize: "dict(norm_obs=True, norm_reward=False)"
+
 # Almost tuned
 BipedalWalker-v3:
   n_envs: 1

hyperparams/ddpg.yml

@@ -131,21 +131,43 @@ HalfCheetah-v3: &mujoco-defaults
   noise_type: 'normal'
   noise_std: 0.1
 
+seals/HalfCheetah-v0:
+  <<: *mujoco-defaults
+
 Ant-v3:
   <<: *mujoco-defaults
 
+seals/Ant-v0:
+  <<: *mujoco-defaults
+
 Hopper-v3:
   <<: *mujoco-defaults
 
+seals/Hopper-v0:
+  <<: *mujoco-defaults
+
 Walker2d-v3:
   <<: *mujoco-defaults
 
+seals/Walker2d-v0:
+  <<: *mujoco-defaults
+
 Humanoid-v3:
   <<: *mujoco-defaults
   n_timesteps: !!float 2e6
 
+seals/Humanoid-v0:
+  <<: *mujoco-defaults
+  n_timesteps: !!float 2e6
+
 Swimmer-v3:
   <<: *mujoco-defaults
   gamma: 0.9999
   train_freq: 1
   gradient_steps: 1
+
+seals/Swimmer-v0:
+  <<: *mujoco-defaults
+  gamma: 0.9999
+  train_freq: 1
+  gradient_steps: 1

hyperparams/dqn.yml

@@ -33,6 +33,21 @@ CartPole-v1:
   exploration_final_eps: 0.04
   policy_kwargs: "dict(net_arch=[256, 256])"
 
+seals/CartPole-v0:
+  n_timesteps: !!float 5e4
+  policy: 'MlpPolicy'
+  learning_rate: !!float 2.3e-3
+  batch_size: 64
+  buffer_size: 100000
+  learning_starts: 1000
+  gamma: 0.99
+  target_update_interval: 10
+  train_freq: 256
+  gradient_steps: 128
+  exploration_fraction: 0.16
+  exploration_final_eps: 0.04
+  policy_kwargs: "dict(net_arch=[256, 256])"
+
 # Tuned
 MountainCar-v0:
   n_timesteps: !!float 1.2e5
@@ -49,6 +64,21 @@ MountainCar-v0:
   exploration_final_eps: 0.07
   policy_kwargs: "dict(net_arch=[256, 256])"
 
+seals/MountainCar-v0:
+  n_timesteps: !!float 1.2e5
+  policy: 'MlpPolicy'
+  learning_rate: !!float 4e-3
+  batch_size: 128
+  buffer_size: 10000
+  learning_starts: 1000
+  gamma: 0.98
+  target_update_interval: 600
+  train_freq: 16
+  gradient_steps: 8
+  exploration_fraction: 0.2
+  exploration_final_eps: 0.07
+  policy_kwargs: "dict(net_arch=[256, 256])"
+
 # Tuned
 LunarLander-v2:
   n_timesteps: !!float 1e5