Correct logging to plot per episode reward rather than per timestep

shariqiqbal2810 · Jul 20, 2020 · 105d60e · 105d60e
1 parent bd263af
commit 105d60e
Showing 1 changed file with 2 additions and 1 deletion.
diff --git a/main.py b/main.py
@@ -98,7 +98,8 @@ def run(config):
         ep_rews = replay_buffer.get_average_rewards(
             config.episode_length * config.n_rollout_threads)
         for a_i, a_ep_rew in enumerate(ep_rews):
-            logger.add_scalar('agent%i/mean_episode_rewards' % a_i, a_ep_rew, ep_i)
+            logger.add_scalar('agent%i/mean_episode_rewards' % a_i,
+                              a_ep_rew * config.episode_length, ep_i)
 
         if ep_i % config.save_interval < config.n_rollout_threads:
             model.prep_rollouts(device='cpu')