wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-01 22:40:06 +08:00
Code Issues Packages Projects Releases Wiki Activity

[rllib] more user-friendly Optimizer signature + compute_apply (#2335)

Browse Source 
* Move signature of optimizers

* fix

* expose compute_apply for policy_graphs

* dictionaries and such

* test for multiagent
This commit is contained in:
Richard Liaw
2018-07-07 12:08:49 -07:00
committed by GitHub
parent e3534c46df
commit e32aed8717
10 changed files with 35 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ray/rllib/optimizers/sync_samples_optimizer.py
+1 -2
View File
@@ -40,8 +40,7 @@ class SyncSamplesOptimizer(PolicyOptimizer):
samples = self.local_evaluator.sample()
with self.grad_timer:
grad, _ = self.local_evaluator.compute_gradients(samples)
self.local_evaluator.apply_gradients(grad)
self.local_evaluator.compute_apply(samples)
self.grad_timer.push_units_processed(samples.count)
self.num_steps_sampled += samples.count