Use max episode length eval in sac evaluation

rlworkgroup · Aug 19, 2020 · ccf66fc · ccf66fc
1 parent bb77db7
commit ccf66fc
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 1 deletion.
diff --git a/examples/torch/sac_half_cheetah_batch.py b/examples/torch/sac_half_cheetah_batch.py
@@ -56,6 +56,7 @@ def sac_half_cheetah_batch(ctxt=None, seed=1):
               qf2=qf2,
               gradient_steps_per_itr=1000,
               max_episode_length=500,
+              max_episode_length_eval=1000,
               replay_buffer=replay_buffer,
               min_buffer_size=1e4,
               target_update_tau=5e-3,

diff --git a/src/garage/torch/algos/sac.py b/src/garage/torch/algos/sac.py
@@ -129,7 +129,8 @@ def __init__(
         self._discount = discount
         self._reward_scale = reward_scale
         self.max_episode_length = max_episode_length
-        self._max_episode_length_eval = max_episode_length_eval
+        self._max_episode_length_eval = (max_episode_length_eval
+                                         or max_episode_length)
 
         self.policy = policy
         self.env_spec = env_spec