From 3c81a15630f93efa094e6d907cd585dbe0a7e072 Mon Sep 17 00:00:00 2001 From: Dominik Roth Date: Tue, 22 Aug 2023 01:18:44 +0200 Subject: [PATCH] Fix: Set initial logging std correctly --- sbBrix/common/policies.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/sbBrix/common/policies.py b/sbBrix/common/policies.py index a01c431..f8f3f4e 100644 --- a/sbBrix/common/policies.py +++ b/sbBrix/common/policies.py @@ -817,7 +817,7 @@ class Actor(BasePolicy): self.mu, self.log_std = self.action_dist.proba_distribution_net( latent_dim=last_layer_dim, return_log_std=True ) - self._remember_log_std = log_std_init + self._remember_log_std = th.Tensor([log_std_init]) # Avoid numerical issues by limiting the mean of the Gaussian # to be in [-clip_mean, clip_mean] if clip_mean > 0.0: