wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-08-20 12:40:44 +08:00
Code Issues Packages Projects Releases Wiki Activity

[rllib] Support the timesteps_per_batch in simple optimizer PPO mode (#2558)

Browse Source 
* support ts

* doc

* Update sync_samples_optimizer.py
This commit is contained in:
Eric Liang
2018-08-06 12:10:59 -07:00
committed by GitHub
parent 9015e742c4
commit 981d9818c1
3 changed files with 17 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ray/rllib/agents/ppo/ppo.py
+4 -2
View File
@@ -83,8 +83,10 @@ class PPOAgent(Agent):
})
if self.config["simple_optimizer"]:
self.optimizer = SyncSamplesOptimizer(
self.local_evaluator, self.remote_evaluators,
{"num_sgd_iter": self.config["num_sgd_iter"]})
self.local_evaluator, self.remote_evaluators, {
"num_sgd_iter": self.config["num_sgd_iter"],
"timesteps_per_batch": self.config["timesteps_per_batch"]
})
else:
self.optimizer = LocalMultiGPUOptimizer(
self.local_evaluator, self.remote_evaluators, {