dodox/metastable-baselines
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fixed SAC+SDE+SDC bugs

Browse Source
This commit is contained in:
dodox
2022-09-03 13:08:31 +02:00
parent 4532135812
commit ee4a0eed56
4 changed files with 25 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
metastable_baselines/distributions/distributions.py
+5 -4
View File
@@ -320,12 +320,12 @@ class UniversalGaussianDistribution(SB3_Distribution):
self.gaussian_actions = mode
return self.prob_squashing_type.apply(mode)
def actions_from_params(self, mean_actions: th.Tensor, log_std: th.Tensor, deterministic: bool = False, latent_pi=None) -> th.Tensor:
def actions_from_params(self, mean_actions: th.Tensor, log_std: th.Tensor, deterministic: bool = False, latent_sde=None) -> th.Tensor:
# Update the proba distribution
self.proba_distribution(mean_actions, log_std, latent_pi=latent_pi)
self.proba_distribution(mean_actions, log_std, latent_sde=latent_sde)
return self.get_actions(deterministic=deterministic)
def log_prob_from_params(self, mean_actions: th.Tensor, log_std: th.Tensor) -> Tuple[th.Tensor, th.Tensor]:
def log_prob_from_params(self, mean_actions: th.Tensor, log_std: th.Tensor, latent_sde=None) -> Tuple[th.Tensor, th.Tensor]:
"""
Compute the log probability of taking an action
given the distribution parameters.
@@ -334,7 +334,8 @@ class UniversalGaussianDistribution(SB3_Distribution):
:param log_std:
:return:
"""
actions = self.actions_from_params(mean_actions, log_std)
actions = self.actions_from_params(
mean_actions, log_std, latent_sde=latent_sde)
log_prob = self.log_prob(actions, self.gaussian_actions)
return actions, log_prob