Update README.md
Browse files
README.md
CHANGED
|
@@ -23,3 +23,18 @@ base_model:
|
|
| 23 |
## 项目团队人员
|
| 24 |
|
| 25 |
齐鲁工业大学（山东省科学院）计算机科学与技术学部 杜宇（[email protected]）
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
## 项目团队人员
|
| 24 |
|
| 25 |
齐鲁工业大学（山东省科学院）计算机科学与技术学部 杜宇（[email protected]）
|
| 26 |
+
|
| 27 |
+
----
|
| 28 |
+
|
| 29 |
+
# Lexipolis-9B-Chat (v1.0) Model Overview
|
| 30 |
+
|
| 31 |
+
**Lexipolis-9B-Chat** version 1.0 is fine-tuned from Tsinghua University's THUDM's `GLM-4-9B-Chat`.
|
| 32 |
+
|
| 33 |
+
It employs the LoRA algorithm for fine-tuning, which includes incremental pre-training and instruction tuning. The incremental training uses a dataset comprising political news, work reports, national laws and regulations, speeches from leaders, political and legal textbooks, official documents, and regulations from national departments, with a data volume of about 1GB.
|
| 34 |
+
|
| 35 |
+
The instruction tuning uses data that includes case trials (aligning case facts with trial results), legal citation (aligning case facts with legal citations), and instruction understanding corpus (aligning instructions for writing "political news" with the content of the news, etc.), with a data volume of about 550MB.
|
| 36 |
+
|
| 37 |
+
## Project Team
|
| 38 |
+
|
| 39 |
+
Faculty of Computer Science and Technology, Qilu University of Technology (Shandong Academy of Sciences)
|
| 40 |
+
Du Yu ([email protected])
|