gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-18 11:40:18 +01:00
Code Issues Projects Releases Wiki Activity

N-step returns for rainbow (#67)

Browse Source 
* n_step returns for rainbow
* Rename CartPole_PPO -> CartPole_ClippedPPO
This commit is contained in:
Gal Leibovich
2018-11-07 18:33:08 +02:00
committed by GitHub
parent 35c477c922
commit 49dea39d34
18 changed files with 173 additions and 117 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
rl_coach/agents/policy_gradients_agent.py
View File

@@ -74,7 +74,7 @@ class PolicyGradientsAgent(PolicyOptimizationAgent):
# batch contains a list of episodes to learn from
network_keys = self.ap.network_wrappers['main'].input_embedders_parameters.keys()
total_returns = batch.total_returns()
total_returns = batch.n_step_discounted_rewards()
for i in reversed(range(batch.size)):
if self.policy_gradient_rescaler == PolicyGradientRescaler.TOTAL_RETURN:
total_returns[i] = total_returns[0]