Want element multiplication, not dot product

sgoodfriend · sgoodfriend · commit ab60ce1a5728 · 2023-04-04T14:56:37.000-07:00
diff --git a/rl_algo_impls/shared/callbacks/microrts_reward_decay_callback.py b/rl_algo_impls/shared/callbacks/microrts_reward_decay_callback.py
@@ -28,7 +28,7 @@ def on_step(self, timesteps_elapsed: int = 1) -> bool:
 
         progress = self.timesteps_elapsed / self.total_train_timesteps
         # Decay all rewards except WinLoss
-        reward_weights = self.base_reward_weights @ np.array(
+        reward_weights = self.base_reward_weights * np.array(
             [1] + [1 - progress] * (len(self.base_reward_weights) - 1)
         )
         self.microrts_env.reward_weight = reward_weights

Original file line number	Diff line number	Diff line change
`@@ -28,7 +28,7 @@ def on_step(self, timesteps_elapsed: int = 1) -> bool:`
`28`	`28`
`29`	`29`	`progress = self.timesteps_elapsed / self.total_train_timesteps`
`30`	`30`	`# Decay all rewards except WinLoss`
`31`		`- reward_weights = self.base_reward_weights @ np.array(`
	`31`	`+ reward_weights = self.base_reward_weights * np.array(`
`32`	`32`	`[1] + [1 - progress] * (len(self.base_reward_weights) - 1)`
`33`	`33`	`)`
`34`	`34`	`self.microrts_env.reward_weight = reward_weights`