Commit ·
7060479
1
Parent(s): ea31a4f
Increase rollouts to 5 for better GRPO learning convergence
Browse files- baseline/evaluator.py +1 -1
baseline/evaluator.py
CHANGED
|
@@ -22,7 +22,7 @@ from core.policy_update import compute_advantage, update_memory
|
|
| 22 |
|
| 23 |
import requests as http_requests
|
| 24 |
|
| 25 |
-
N_ROLLOUTS =
|
| 26 |
|
| 27 |
|
| 28 |
# ── Prompt Builder With Memory ────────────────────────────────────────────────
|
|
|
|
| 22 |
|
| 23 |
import requests as http_requests
|
| 24 |
|
| 25 |
+
N_ROLLOUTS = 5
|
| 26 |
|
| 27 |
|
| 28 |
# ── Prompt Builder With Memory ────────────────────────────────────────────────
|