wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-11 16:34:48 +08:00
Code Issues Packages Projects Releases Wiki Activity

[rllib] Update RLlib to work with new actor scheduling behavior (#1754)

Browse Source 
* Mon Mar 19 21:23:01 PDT 2018

* Mon Mar 19 21:23:07 PDT 2018

* Mon Mar 19 21:30:49 PDT 2018

* Mon Mar 19 21:32:05 PDT 2018

* Mon Mar 19 21:35:43 PDT 2018

* fix cpu limits

* Mon Mar 19 22:25:07 PDT 2018
This commit is contained in:
Eric Liang
2018-03-20 19:29:52 -07:00
committed by GitHub
parent 4bccabd910
commit b41bdcefa0
12 changed files with 31 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ray/rllib/optimizers/apex_optimizer.py
+10 -7
View File
@@ -18,7 +18,7 @@ import ray
from ray.rllib.optimizers.policy_optimizer import PolicyOptimizer
from ray.rllib.optimizers.replay_buffer import PrioritizedReplayBuffer
from ray.rllib.optimizers.sample_batch import SampleBatch
from ray.rllib.utils.actors import TaskPool, create_colocated
from ray.rllib.utils.actors import TaskPool
from ray.rllib.utils.timer import TimerStat
from ray.rllib.utils.window_stat import WindowStat
@@ -163,12 +163,15 @@ class ApexOptimizer(PolicyOptimizer):
self.learner = LearnerThread(self.local_evaluator)
self.learner.start()
self.replay_actors = create_colocated(
ReplayActor,
[num_replay_buffer_shards, learning_starts, buffer_size,
train_batch_size, prioritized_replay_alpha,
prioritized_replay_beta, prioritized_replay_eps, clip_rewards],
num_replay_buffer_shards)
# TODO(ekl) use create_colocated() for these actors once
# https://github.com/ray-project/ray/issues/1734 is fixed
self.replay_actors = [
ReplayActor.remote(
num_replay_buffer_shards, learning_starts, buffer_size,
train_batch_size, prioritized_replay_alpha,
prioritized_replay_beta, prioritized_replay_eps, clip_rewards)
for _ in range(num_replay_buffer_shards)
]
assert len(self.remote_evaluators) > 0
# Stats