Skip to content

Commit

Permalink
Added experiment specification examples
Browse files Browse the repository at this point in the history
  • Loading branch information
@arnomoonens
arnomoonens committed Mar 15, 2018
1 parent 7155abe commit a7b98f4
Show file tree
Hide file tree
Showing 20 changed files with 373 additions and 8 deletions.
2 changes: 1 addition & 1 deletion .gitignore
View file
Original file line number Diff line number Diff line change
Expand Up @@ -93,7 +93,7 @@ ENV/
/DeepRL.sublime-workspace
Environment/*.json
*.orig
*-experiment.json
/*-experiment.json
*~
.*.swp
.vscode/
Expand Down
6 changes: 3 additions & 3 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -41,15 +41,15 @@ pip install -r requirements.txt

### Algorithms/experiments

You can run algorithms by passing an experiment specification (in _json_ format) to `main.py`:
You can run algorithms by passing the path to an experiment specification (which is a file in _json_ format) to `main.py`:

```Shell

python main.py <experiment_description>
python main.py <path_to_experiment_specification>

```

[Example of an experiment specification](./experiment_spec_example.json)
Examples of experiment specifications can be found in the [_experiment_specs_](./experiment_specs) folder.

### Statistics

Expand Down
16 changes: 16 additions & 0 deletions experiment_specs/CartPole-v0-A2C-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,16 @@
{
"experiment_name": "CartPole-CEM",
"environments": {
"type": "single",
"source": "CartPole-v0"
},
"agent": {
"name": "REINFORCE",
"args": {
"monitor_path": "/tmp/CartPole-v0-RE",
"video": false,
"save_model": true,
"monitor": true
}
}
}
27 changes: 27 additions & 0 deletions experiment_specs/DPPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,27 @@
{
"experiment_name": "CartPole-DPPO",
"environments": {
"type": "single",
"source": "CartPole-v0"
},
"agent": {
"name": "DPPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-DPPO",
"n_hidden_units": 10,
"n_hidden_layers": 1,
"gradient_clip_value": 50.0,
"batch_size": 64,
"learning_rate": 3e-4,
"vf_coef": 1.0,
"n_local_steps": 512,
"n_workers": 4,
"cso_epsilon": 0.2,
"n_epochs": 10,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
27 changes: 27 additions & 0 deletions experiment_specs/DTRPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,27 @@
{
"experiment_name": "CartPole-DPPO",
"environments": {
"type": "single",
"source": "CartPole-v0"
},
"agent": {
"name": "TRPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-DPPO",
"n_hidden_units": 10,
"n_hidden_layers": 1,
"gradient_clip_value": 50.0,
"batch_size": 64,
"learning_rate": 3e-4,
"vf_coef": 1.0,
"n_local_steps": 512,
"n_workers": 4,
"cso_epsilon": 0.2,
"n_epochs": 10,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
27 changes: 27 additions & 0 deletions experiment_specs/FrozenLake-DPPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,27 @@
{
"experiment_name": "CartPole-DPPO",
"environments": {
"type": "single",
"source": "FrozenLake8x8-v0"
},
"agent": {
"name": "DPPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-DPPO",
"n_hidden_units": 10,
"n_hidden_layers": 1,
"gradient_clip_value": 50.0,
"batch_size": 64,
"learning_rate": 3e-4,
"vf_coef": 1.0,
"n_local_steps": 512,
"n_workers": 4,
"cso_epsilon": 0.2,
"n_epochs": 10,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
23 changes: 23 additions & 0 deletions experiment_specs/FrozenLake-PPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
{
"experiment_name": "CartPole-PPO",
"environments": {
"type": "single",
"source": "FrozenLake8x8-v0"
},
"agent": {
"name": "PPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-PPO",
"n_hidden": 100,
"gradient_clip_value": 50.0,
"batch_size": 256,
"learning_rate": 3e-4,
"vf_coef": 1.0,
"n_local_steps": 256,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
8 changes: 4 additions & 4 deletions experiment_spec_example.json → ...cs/MountainCar-v0-SarsaFA-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -1,16 +1,16 @@
{
"experiment_name": "CartPole-SarsaFA",
"experiment_name": "MountainCar-SarsaFA",
"environments": {
"type": "single",
"source": "MountainCar-v0"
},
"agent": {
"name": "SarsaFA",
"args": {
"monitor_path": "/tmp/CartPole-v0-SarsaFA",
"video": false,
"monitor_path": "/tmp/MountainCar-SarsaFA",
"video": true,
"save_model": true,
"monitor": true
}
}
}
}
23 changes: 23 additions & 0 deletions experiment_specs/PPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
{
"experiment_name": "CartPole-PPO",
"environments": {
"type": "single",
"source": "CartPole-v0"
},
"agent": {
"name": "PPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-PPO",
"n_hidden": 10,
"gradient_clip_value": 50.0,
"batch_size": 64,
"learning_rate": 3e-4,
"vf_coef": 1.0,
"n_local_steps": 2048,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
20 changes: 20 additions & 0 deletions experiment_specs/Pendulum-A3C-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
{
"experiment_name": "CartPole-A3C",
"environments": {
"type": "single",
"source": "Pendulum-v0"
},
"agent": {
"name": "A3C",
"args": {
"monitor_path": "/tmp/CartPole-v0-A3C",
"n_tasks": 2,
"video": false,
"save_model": false,
"monitor": true,
"T_max": 100000,
"n_iter": 1000,
"shared_optimizer": true
}
}
}
18 changes: 18 additions & 0 deletions experiment_specs/Pendulum-DDPG-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
{
"experiment_name": "CartPole-PPO",
"environments": {
"type": "single",
"source": "Pendulum-v0"
},
"agent": {
"name": "DDPG",
"args": {
"monitor_path": "/tmp/CartPole-v0-PPO",
"video": false,
"n_hidden_units": 64,
"save_model": true,
"monitor": true,
"n_episodes": 400000
}
}
}
24 changes: 24 additions & 0 deletions experiment_specs/Pendulum-PPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,24 @@
{
"experiment_name": "CartPole-PPO",
"environments": {
"type": "single",
"source": "Pendulum-v0"
},
"agent": {
"name": "PPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-PPO",
"n_hidden_units": 20,
"n_hidden_layers": 2,
"batch_size": 64,
"learning_rate": 3e-4,
"vf_coef": 1.0,
"entropy_coef": 0.0,
"n_local_steps": 2048,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
27 changes: 27 additions & 0 deletions experiment_specs/Pendulum-v0-DPPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,27 @@
{
"experiment_name": "CartPole-DPPO",
"environments": {
"type": "single",
"source": "Pendulum-v0"
},
"agent": {
"name": "DPPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-DPPO",
"n_hidden_units": 10,
"n_hidden_layers": 1,
"gradient_clip_value": 50.0,
"batch_size": 64,
"learning_rate": 3e-4,
"vf_coef": 1.0,
"n_local_steps": 512,
"n_workers": 4,
"cso_epsilon": 0.2,
"n_epochs": 10,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
18 changes: 18 additions & 0 deletions experiment_specs/PongDeterministic-v4-A2C-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
{
"experiment_name": "Pong-REINFORCE",
"environments": {
"type": "single",
"source": "PongDeterministic-v4"
},
"agent": {
"name": "A2C",
"args": {
"monitor_path": "/tmp/CartPole-v0-KT",
"RNN": true,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 800000
}
}
}
20 changes: 20 additions & 0 deletions experiment_specs/PongDeterministic-v4-A3C-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
{
"experiment_name": "Pong-REINFORCE",
"environments": {
"type": "single",
"source": "PongDeterministic-v4"
},
"agent": {
"name": "A3C",
"args": {
"monitor_path": "/tmp/CartPole-v0-KT",
"RNN": true,
"video": false,
"save_model": true,
"monitor": true,
"T_max": 8000,
"n_tasks": 1,
"shared_optimizer": false
}
}
}
18 changes: 18 additions & 0 deletions experiment_specs/PongDeterministic-v4-DPPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
{
"experiment_name": "CartPole-A2C",
"environments": {
"type": "single",
"source": "PongDeterministic-v4"
},
"agent": {
"name": "DPPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-A2C",
"gradient_clip_value": 1.0,
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 400000
}
}
}
18 changes: 18 additions & 0 deletions experiment_specs/PongDeterministic-v4-PPO-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
{
"experiment_name": "Pong-REINFORCE",
"environments": {
"type": "single",
"source": "PongDeterministic-v4"
},
"agent": {
"name": "PPO",
"args": {
"monitor_path": "/tmp/CartPole-v0-KT",
"RNN": true,
"video": false,
"save_model": true,
"monitor": true,
"T_max": 8000
}
}
}
19 changes: 19 additions & 0 deletions experiment_specs/PongDeterministic-v4-RE-experiment.json
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,19 @@
{
"experiment_name": "CartPole-RE",
"environments": {
"type": "single",
"source": "PongDeterministic-v4"
},
"agent": {
"name": "REINFORCE",
"args": {
"RNN": false,
"monitor_path": "/tmp/CartPole-v0-RE",
"video": false,
"save_model": true,
"monitor": true,
"n_iter": 200,
"switch_at_iter": 10
}
}
}
Loading

0 comments on commit a7b98f4

Please sign in to comment.