fixes
This commit is contained in:
parent
5739f365ab
commit
d6c05668cb
@ -817,6 +817,7 @@ class Actor(BasePolicy):
|
|||||||
self.mu, self.log_std = self.action_dist.proba_distribution_net(
|
self.mu, self.log_std = self.action_dist.proba_distribution_net(
|
||||||
latent_dim=last_layer_dim, return_log_std=True, **dist_kwargs
|
latent_dim=last_layer_dim, return_log_std=True, **dist_kwargs
|
||||||
)
|
)
|
||||||
|
self._remember_log_std = log_std_init
|
||||||
# Avoid numerical issues by limiting the mean of the Gaussian
|
# Avoid numerical issues by limiting the mean of the Gaussian
|
||||||
# to be in [-clip_mean, clip_mean]
|
# to be in [-clip_mean, clip_mean]
|
||||||
if clip_mean > 0.0:
|
if clip_mean > 0.0:
|
||||||
|
Loading…
Reference in New Issue
Block a user