Commit
•
7ea512c
1
Parent(s):
e897962
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,17 @@
|
|
1 |
---
|
2 |
-
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
library_name: peft
|
3 |
+
tags:
|
4 |
+
- text-to-sql
|
5 |
+
- spider
|
6 |
+
- 'text2sql'
|
7 |
+
|
8 |
---
|
9 |
+
## Introduce
|
10 |
+
This folder is a text-to-sql weights directory, containing weight files fine-tuned based on the LoRA with the CodeLlama-13b-Instruct-hf model through the [DB-GPT-Hub](https://github.com/eosphoros-ai/DB-GPT-Hub/tree/main) project. The training data used is from the Spider training set. This weights files achieving an execution accuracy of approximately 0.789 on the Spider evaluation set.
|
11 |
+
Merge the weights with [CodeLlama-13b-Instruct-hf](https://huggingface.co/codellama/CodeLlama-13b-Instruct-hf/tree/main) and this folder weigths, you can refer the [DB-GPT-Hub](https://github.com/eosphoros-ai/DB-GPT-Hub/tree/main) ,in `dbgpt_hub/scripts/export_merge.sh`.
|
12 |
+
If you find our weight files or the DB-GPT-Hub project helpful for your work, give a star on our github project [DB-GPT-Hub](https://github.com/eosphoros-ai/DB-GPT-Hub/tree/main) will be a great encouragement for us to release more weight files.
|
13 |
+
|
14 |
+
### Framework versions
|
15 |
+
|
16 |
+
|
17 |
+
- PEFT 0.4.0
|