saymrwulf/stable-baselines3
1
0
Fork 0
mirror of https://github.com/saymrwulf/stable-baselines3.git synced 2026-07-30 20:18:15 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Entropy can't be None

Browse source
This commit is contained in:
Quentin Gallouédec 2022-11-29 13:10:05 +01:00
parent 838165c265
commit 6da012f9ab
1 changed files with 5 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
stable_baselines3/common/distributions.py
View file

@ -54,7 +54,7 @@ class Distribution(ABC):
"""
@abstractmethod
def entropy(self) -> Optional[th.Tensor]:
def entropy(self) -> th.Tensor:
"""
Returns Shannon's entropy of the probability
@ -240,10 +240,8 @@ class SquashedDiagGaussianDistribution(DiagGaussianDistribution):
log_prob -= th.sum(th.log(1 - actions**2 + self.epsilon), dim=1)
return log_prob
def entropy(self) -> Optional[th.Tensor]:
# No analytical form,
# entropy needs to be estimated using -log_prob.mean()
return None
def entropy(self) -> th.Tensor:
raise NotImplementedError("No analytical form, entropy needs to be estimated using -log_prob.mean()")
def sample(self) -> th.Tensor:
# Reparametrization trick to pass gradients
@ -567,11 +565,9 @@ class StateDependentNoiseDistribution(Distribution):
log_prob -= th.sum(self.bijector.log_prob_correction(gaussian_actions), dim=1)
return log_prob
def entropy(self) -> Optional[th.Tensor]:
def entropy(self) -> th.Tensor:
if self.bijector is not None:
# No analytical form,
# entropy needs to be estimated using -log_prob.mean()
return None
raise NotImplementedError("No analytical form, entropy needs to be estimated using -log_prob.mean()")
return sum_independent_dims(self.distribution.entropy())
def sample(self) -> th.Tensor: