diff --git a/torchrl/objectives/dqn.py b/torchrl/objectives/dqn.py index a9d50cadd50..6cbb8b02426 100644 --- a/torchrl/objectives/dqn.py +++ b/torchrl/objectives/dqn.py @@ -47,7 +47,7 @@ class DQNLoss(LossModule): Defaults to "l2". delay_value (bool, optional): whether to duplicate the value network into a new target value network to - create a DQN with a target network. Default is ``False``. + create a DQN with a target network. Default is ``True``. double_dqn (bool, optional): whether to use Double DQN, as described in https://arxiv.org/abs/1509.06461. Defaults to ``False``. action_space (str or TensorSpec, optional): Action space. Must be one of diff --git a/tutorials/sphinx-tutorials/getting-started-5.py b/tutorials/sphinx-tutorials/getting-started-5.py index 5f95fe1e534..d355d1888c5 100644 --- a/tutorials/sphinx-tutorials/getting-started-5.py +++ b/tutorials/sphinx-tutorials/getting-started-5.py @@ -89,7 +89,7 @@ optim_steps = 10 collector = SyncDataCollector( env, - policy, + policy_explore, frames_per_batch=frames_per_batch, total_frames=-1, init_random_frames=init_rand_steps,