From 826b2250f430f02ce5afd709af8d18c80f1bf659 Mon Sep 17 00:00:00 2001 From: Antonin RAFFIN Date: Sat, 29 Aug 2020 17:48:31 +0200 Subject: [PATCH] disable pytype for n-step replay --- stable_baselines3/common/buffers.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/stable_baselines3/common/buffers.py b/stable_baselines3/common/buffers.py index 60dee19..64cba78 100644 --- a/stable_baselines3/common/buffers.py +++ b/stable_baselines3/common/buffers.py @@ -463,7 +463,7 @@ class NstepReplayBuffer(ReplayBuffer): # Memory inneficient version but fast computation self.log_probs[unique_indices] = log_prob.cpu().numpy().flatten() # Add entropy term, only for n-step > 1 - rewards[:, 1:] = rewards[:, 1:] - self.ent_coef * self.log_probs[indices[:, 1:]] + rewards[:, 1:] = rewards[:, 1:] - self.ent_coef * self.log_probs[indices[:, 1:]] # pytype: disable=attribute-error # we filter through the indices. # The immediate indice, i.e. col 0 needs to be 1, so we ensure that it is here using np.ones