[RLlib; Off-policy] Add episode sampling to EpisodeReplayBuffer. (r…

…ay-project#47500) Signed-off-by: ujjawal-khare <ujjawal.khare@dream11.com>
ujjawal-khare-27 · Oct 15, 2024 · 882510b · 882510b
1 parent 3dca501
commit 882510b
Show file tree

Hide file tree

Showing 2 changed files with 0 additions and 8 deletions.
diff --git a/rllib/utils/replay_buffers/episode_replay_buffer.py b/rllib/utils/replay_buffers/episode_replay_buffer.py
@@ -218,7 +218,6 @@ def sample(
         include_infos: bool = False,
         include_extra_model_outputs: bool = False,
         sample_episodes: Optional[bool] = False,
-        finalize: bool = False,
         **kwargs,
     ) -> Union[SampleBatchType, SingleAgentEpisode]:
         """Samples from a buffer in a randomized way.
@@ -263,7 +262,6 @@ def sample(
                 actual state of model e.g. action log-probabilities, etc.). If `True`,
                 the extra model outputs at the `"obs"` in the batch is included (the
                 timestep at which the action is computed).
-            finalize: If episodes should be finalized.
 
         Returns:
             Either a batch with transitions in each row or (if `return_episodes=True`)
@@ -281,7 +279,6 @@ def sample(
                 gamma=gamma,
                 include_infos=include_infos,
                 include_extra_model_outputs=include_extra_model_outputs,
-                finalize=finalize,
             )
         else:
             return self._sample_batch(
@@ -427,7 +424,6 @@ def _sample_episodes(
         gamma: float = 0.99,
         include_infos: bool = False,
         include_extra_model_outputs: bool = False,
-        finalize: bool = False,
         **kwargs,
     ) -> List[SingleAgentEpisode]:
         """Samples episodes from a buffer in a randomized way.
@@ -472,7 +468,6 @@ def _sample_episodes(
                 actual state of model e.g. action log-probabilities, etc.). If `True`,
                 the extra model outputs at the `"obs"` in the batch is included (the
                 timestep at which the action is computed).
-            finalize: If episodes should be finalized.
 
         Returns:
             A list of 1-step long episodes containing all basic episode data and if
@@ -583,8 +578,6 @@ def _sample_episodes(
                 len_lookback_buffer=0,
                 t_started=episode_ts,
             )
-            if finalize:
-                sampled_episode.finalize()
             sampled_episodes.append(sampled_episode)
 
             # Increment counter.

diff --git a/rllib/utils/replay_buffers/prioritized_episode_buffer.py b/rllib/utils/replay_buffers/prioritized_episode_buffer.py
@@ -310,7 +310,6 @@ def sample(
         gamma: float = 0.99,
         include_infos: bool = False,
         include_extra_model_outputs: bool = False,
-        finalize: bool = False,
         **kwargs,
     ) -> SampleBatchType:
         """Samples from a buffer in a prioritized way.