Update README.md
Browse files
README.md
CHANGED
|
@@ -33,6 +33,22 @@ AI 와 빅데이터 분석 전문 기업인 Linkbricks의 데이터사이언티
|
|
| 33 |
-Deepspeed Stage=3, rslora 및 BAdam Layer Mode 사용<br>
|
| 34 |
-ollama run benedict/linkbricks-gemma2-korean:27b
|
| 35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
Dr. Yunsung Ji (Saxo), a data scientist at Linkbricks, a company specializing in AI and big data analytics, fine-tuned the gemma-2-27b-it base model with SFT->DPO using four H100-80Gs on KT-CLOUD.
|
| 37 |
It is a Korean language model trained to handle complex Korean logic problems through Korean-Chinese-English-Japanese cross-training data and logical data, and Tokenizer uses the base model without word expansion.
|
| 38 |
|
|
|
|
| 33 |
-Deepspeed Stage=3, rslora 및 BAdam Layer Mode 사용<br>
|
| 34 |
-ollama run benedict/linkbricks-gemma2-korean:27b
|
| 35 |
|
| 36 |
+
<br><br>
|
| 37 |
+
Benchmark (Open Ko LLM Leader Board Season 2 : No. 1)<br>
|
| 38 |
+
Model : Saxo/Linkbricks-Horizon-AI-Korean-Gemma-2-sft-dpo-27B<br>
|
| 39 |
+
Average : 51.37<br>
|
| 40 |
+
Ko-GPQA : 25.25<br>
|
| 41 |
+
Ko-Winogrande : 68.27<br>
|
| 42 |
+
Ko-GSM8k : 70.96<br>
|
| 43 |
+
Ko-EQ Bench : 50.25<br>
|
| 44 |
+
Ko-IFEval : 49.84<br>
|
| 45 |
+
KorNAT-CKA : 34.59<br>
|
| 46 |
+
KorNAT-SVA : 48.42<br>
|
| 47 |
+
Ko-Harmlessness : 65.66<br>
|
| 48 |
+
Ko-Helpfulness : 49.12<br>
|
| 49 |
+
|
| 50 |
+
<br><br>
|
| 51 |
+
|
| 52 |
Dr. Yunsung Ji (Saxo), a data scientist at Linkbricks, a company specializing in AI and big data analytics, fine-tuned the gemma-2-27b-it base model with SFT->DPO using four H100-80Gs on KT-CLOUD.
|
| 53 |
It is a Korean language model trained to handle complex Korean logic problems through Korean-Chinese-English-Japanese cross-training data and logical data, and Tokenizer uses the base model without word expansion.
|
| 54 |
|