huangfu/InteractionImitation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix tune by moving file

Browse Source
This commit is contained in:
ebuehrle
2022-02-23 17:32:11 +01:00
parent f037c119cc
commit 406c4ad9ee
2 changed files with 148 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

62
scratch/etienne/trpo/experiments/sgail-ppo-options-setobs2.py
View File

@@ -17,38 +17,38 @@ from src.safe_options.options import SafeOptionsEnv
from torch.utils.tensorboard import SummaryWriter
from ray import tune
obs_min = np.array([
[-1000, -1000, 0, -np.pi, -1e-1, 0.],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
]).reshape(-1)
obs_max = np.array([
[1000, 1000, 20, np.pi, 1e-1, 0.],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
]).reshape(-1)
envs = [SafeOptionsEnv(Setobs(
TransformObservation(CollisionPenaltyWrapper(IntersimpleLidarFlatRandom(
n_rays=5,
reward=functools.partial(
speed_reward,
collision_penalty=0
),
stop_on_collision=True,
), collision_distance=6, collision_penalty=100), lambda obs: (obs - obs_min) / (obs_max - obs_min + 1e-10))
), options=[(0, 5), (1, 5), (2, 5), (4, 5), (6, 5), (8, 5), (10, 5)], safe_actions_collision_method='circle', abort_unsafe_collision_method='circle') for _ in range(60)]
env_fn = lambda i: envs[i]
def training_function(config):
obs_min = np.array([
[-1000, -1000, 0, -np.pi, -1e-1, 0.],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
[0, -np.pi, -20, -20, -np.pi, -1e-1],
]).reshape(-1)
obs_max = np.array([
[1000, 1000, 20, np.pi, 1e-1, 0.],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
[50, np.pi, 20, 20, np.pi, 1e-1],
]).reshape(-1)
envs = [SafeOptionsEnv(Setobs(
TransformObservation(CollisionPenaltyWrapper(IntersimpleLidarFlatRandom(
n_rays=5,
reward=functools.partial(
speed_reward,
collision_penalty=0
),
stop_on_collision=True,
), collision_distance=6, collision_penalty=100), lambda obs: (obs - obs_min) / (obs_max - obs_min + 1e-10))
), options=[(0, 5), (1, 5), (2, 5), (4, 5), (6, 5), (8, 5), (10, 5)], safe_actions_collision_method='circle', abort_unsafe_collision_method='circle') for _ in range(60)]
env_fn = lambda i: envs[i]
policy = SetMaskedDiscretePolicy(env_fn(0).action_space.n) # config net architecture
pi_opt = torch.optim.Adam(policy.parameters(), lr=3e-5) # config learning rate
pi_lr_scheduler = torch.optim.lr_scheduler.ExponentialLR(pi_opt, gamma=0.98) # config lr decay