From 504da45e699cae4f8e62c6fa39ab7d504706d356 Mon Sep 17 00:00:00 2001 From: Thomas Lecat Date: Fri, 11 Sep 2020 00:03:02 +0200 Subject: [PATCH] fix(rllib): allow explore=False with tuple action distributions (#10443) --- rllib/utils/exploration/stochastic_sampling.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/rllib/utils/exploration/stochastic_sampling.py b/rllib/utils/exploration/stochastic_sampling.py index 8d188a735..9ec27bd68 100644 --- a/rllib/utils/exploration/stochastic_sampling.py +++ b/rllib/utils/exploration/stochastic_sampling.py @@ -72,5 +72,5 @@ class StochasticSampling(Exploration): logp = action_dist.sampled_action_logp() else: action = action_dist.deterministic_sample() - logp = torch.zeros((action.size()[0], ), dtype=torch.float32) + logp = torch.zeros_like(action_dist.sampled_action_logp()) return action, logp