diff --git a/rllib/models/torch/torch_action_dist.py b/rllib/models/torch/torch_action_dist.py index 78c1c4348..9605f6e3c 100644 --- a/rllib/models/torch/torch_action_dist.py +++ b/rllib/models/torch/torch_action_dist.py @@ -308,7 +308,7 @@ class TorchDeterministic(TorchDistributionWrapper): @override(TorchDistributionWrapper) def sampled_action_logp(self): - return 0.0 + return torch.zeros((self.inputs.size()[0], ), dtype=torch.float32) @override(TorchDistributionWrapper) def sample(self):