saymrwulf/stable-baselines3
1
0
Fork 0
mirror of https://github.com/saymrwulf/stable-baselines3.git synced 2026-07-05 04:17:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

evaluation_actions returns approx entropy when no analytical form is known

Browse source
This commit is contained in:
Quentin Gallouédec 2022-11-29 15:26:10 +01:00
parent f262cac7ba
commit beeea59d9b
1 changed files with 3 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
stable_baselines3/common/policies.py
View file

@ -612,7 +612,7 @@ class ActorCriticPolicy(BasePolicy):
"""
return self.get_distribution(observation).get_actions(deterministic=deterministic)
def evaluate_actions(self, obs: th.Tensor, actions: th.Tensor) -> Tuple[th.Tensor, th.Tensor, Optional[th.Tensor]]:
def evaluate_actions(self, obs: th.Tensor, actions: th.Tensor) -> Tuple[th.Tensor, th.Tensor, th.Tensor]:
"""
Evaluate actions according to the current policy,
given the observations.
@ -631,7 +631,8 @@ class ActorCriticPolicy(BasePolicy):
try:
entropy = distribution.entropy()
except NotImplementedError:
entropy = None
# When no analytical form, entropy needs to be estimated using -log_prob.mean()
entropy = -log_prob.mean(dim=1)
return values, log_prob, entropy
def get_distribution(self, obs: th.Tensor) -> Distribution: