-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
chore(exp): add SAC extra experiments (#426)
This commit adds the extra experiments for the SAC algorithm in the cartpole, oscillator and comp oscillator environments.
- Loading branch information
Showing
118 changed files
with
3,084 additions
and
30 deletions.
There are no files selected for viewing
42 changes: 42 additions & 0 deletions
42
...ha/han2020_reproduction_sac_cartpole_cost_alpha3_tune_experiment_bigger_initial_alpha.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_bigger_initial_alpha | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 50 # NOTE: Decreased to 50 for consistency with the codebase. | ||
num_test_episodes: 10 | ||
alpha: 2.0 | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 78456 48104 567 3658 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_bigger_initial_alpha" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
...20_reproduction_sac_cartpole_cost_alpha3_tune_experiment_seed234_bigger_initial_alpha.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_bigger_initial_alpha | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 50 # NOTE: Decreased to 50 for consistency with the codebase. | ||
num_test_episodes: 10 | ||
alpha: 2.0 | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_bigger_initial_alpha" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
...0_reproduction_sac_cartpole_cost_alpha3_tune_experiment_seed3658_bigger_initial_alpha.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_bigger_initial_alpha | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 50 # NOTE: Decreased to 50 for consistency with the codebase. | ||
num_test_episodes: 10 | ||
alpha: 2.0 | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 3658 # 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_bigger_initial_alpha" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
..._reproduction_sac_cartpole_cost_alpha3_tune_experiment_seed48104_bigger_initial_alpha.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_bigger_initial_alpha | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 50 # NOTE: Decreased to 50 for consistency with the codebase. | ||
num_test_episodes: 10 | ||
alpha: 2.0 | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 48104 # 567 3658 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_bigger_initial_alpha" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
...20_reproduction_sac_cartpole_cost_alpha3_tune_experiment_seed567_bigger_initial_alpha.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_bigger_initial_alpha | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 50 # NOTE: Decreased to 50 for consistency with the codebase. | ||
num_test_episodes: 10 | ||
alpha: 2.0 | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 567 # 3658 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_bigger_initial_alpha" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
..._reproduction_sac_cartpole_cost_alpha3_tune_experiment_seed78456_bigger_initial_alpha.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_bigger_initial_alpha | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 50 # NOTE: Decreased to 50 for consistency with the codebase. | ||
num_test_episodes: 10 | ||
alpha: 2.0 | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 78456 # 48104 567 3658 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_bigger_initial_alpha" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
...2020_reproduction_sac_cartpole_cost_alpha3_tune_experiment_different_steps_per_update.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_different_steps_per_update | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 80 | ||
num_test_episodes: 10 | ||
alpha: 1.0 # NOTE: Decreased to 1.0 for consistency with the codebase. | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 78456 48104 567 3658 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_different_steps_per_update" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
...roduction_sac_cartpole_cost_alpha3_tune_experiment_seed234_different_steps_per_update.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_different_steps_per_update | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 80 | ||
num_test_episodes: 10 | ||
alpha: 1.0 # NOTE: Decreased to 1.0 for consistency with the codebase. | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_different_steps_per_update" | ||
device: "gpu:1" |
42 changes: 42 additions & 0 deletions
42
...oduction_sac_cartpole_cost_alpha3_tune_experiment_seed3658_different_steps_per_update.yml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
alg_name: sac | ||
exp_name: han2020_reproduction_sac_cartpole_cost_alpha3_tune_exp_different_steps_per_update | ||
env_name: "stable_gym:CartPoleCost-v1" | ||
ac_kwargs: | ||
hidden_sizes: | ||
actor: [256, 256] # NOTE: Using [256, 256] for consistency with the article. | ||
critic: [256, 256] # NOTE: Use [256, 256] for consistency with the codebase. | ||
activation: | ||
actor: "nn.ReLU" | ||
critic: "nn.ReLU" | ||
output_activation: | ||
actor: "nn.ReLU" | ||
opt_type: "minimize" | ||
max_ep_len: 250 | ||
epochs: 489 | ||
steps_per_epoch: 2048 | ||
start_steps: 0 | ||
update_every: 100 | ||
update_after: 1000 | ||
steps_per_update: 80 | ||
num_test_episodes: 10 | ||
alpha: 1.0 # NOTE: Decreased to 1.0 for consistency with the codebase. | ||
gamma: 0.995 | ||
polyak: 0.995 | ||
adaptive_temperature: True | ||
lr_a: "1e-4" | ||
lr_c: "3e-4" | ||
lr_alpha: "1e-4" | ||
lr_a_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_c_final: "3e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_alpha_final: "1e-10" # NOTE: Not in Han et al. 2020, but aligned with original code. | ||
lr_a_decay_type: "linear" | ||
lr_c_decay_type: "linear" | ||
lr_alpha_decay_type: "linear" | ||
lr_decay_ref: "step" | ||
batch_size: 256 | ||
replay_size: "int(1e6)" | ||
seed: 3658 # 234 # NOTE: Using 5 seeds for tuning. | ||
save_freq: 10 | ||
use_wandb: True | ||
wandb_group: "han2020_reproduction_sac_different_steps_per_update" | ||
device: "gpu:1" |
Oops, something went wrong.