Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,3 +1,23 @@
|
|
| 1 |
-
---
|
| 2 |
-
license:
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- trl
|
| 5 |
+
- ppo
|
| 6 |
+
- lora
|
| 7 |
+
- alignment
|
| 8 |
+
- reward-modeling
|
| 9 |
+
- ultrafeedback
|
| 10 |
+
base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# Aligned TinyLlama on UltraFeedback (fixed-1k prompt pool)
|
| 14 |
+
|
| 15 |
+
This model was aligned with **TRL PPO** using a reward model:
|
| 16 |
+
- **payelb/UltraFeedback_openbmb_deberta_1k_fixed_WoN** (tag: `won`)
|
| 17 |
+
|
| 18 |
+
Key settings:
|
| 19 |
+
- Prompt pool: restricted to the same fixed/selected 1k subset used for RM training (loaded from CSV)
|
| 20 |
+
- PPO updates: 200
|
| 21 |
+
- batch size: 4
|
| 22 |
+
- lr: 1e-05
|
| 23 |
+
- LoRA: r=16, alpha=32, dropout=0.05
|