Spaces:
Sleeping
Sleeping
Commit ·
d5dbfe8
1
Parent(s): 5961585
Clamp shaped rewards to non-negative values
Browse filesCo-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
backend/src/polypharmacy_env/rewards.py
CHANGED
|
@@ -93,4 +93,4 @@ def compute_shaped_reward(
|
|
| 93 |
|
| 94 |
# finish_review terminal bonus is added by the caller after grading
|
| 95 |
|
| 96 |
-
return reward
|
|
|
|
| 93 |
|
| 94 |
# finish_review terminal bonus is added by the caller after grading
|
| 95 |
|
| 96 |
+
return max(0.0, reward)
|