Upload phase1_summary.txt with huggingface_hub
Browse files- phase1_summary.txt +17 -0
phase1_summary.txt
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
=== Phase 1 Summary ===
|
| 2 |
+
|
| 3 |
+
Training: MaskablePPO vs Random Opponents
|
| 4 |
+
Timesteps: 500,352
|
| 5 |
+
Final Training Reward: 237.0
|
| 6 |
+
|
| 7 |
+
Evaluation (100 episodes vs Random):
|
| 8 |
+
=== TIL-26-AE Phase 1 Evaluation Results ===
|
| 9 |
+
Model: phase1_final.zip (500k steps)
|
| 10 |
+
Episodes: 100
|
| 11 |
+
Win Rate: 92.0% (92/100)
|
| 12 |
+
Avg Reward: 180.1
|
| 13 |
+
Avg Episode Length: 200.0
|
| 14 |
+
Avg Bombs/Episode: 20.4
|
| 15 |
+
Survival Rate (198+ steps): 100.0%
|
| 16 |
+
|
| 17 |
+
Checkpoints saved: ckpt_50000 to ckpt_400000 + phase1_final.zip
|