Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: en
|
3 |
+
widget:
|
4 |
+
- text: >-
|
5 |
+
convert question and table into SQL query. tables: people_name(id,name),
|
6 |
+
people_age(people_id,age). question: how many people with name jui and age
|
7 |
+
less than 25
|
8 |
+
license: cc-by-sa-4.0
|
9 |
+
---
|
10 |
+
|
11 |
+
This is an upgraded version of [https://huggingface.co/juierror/flan-t5-text2sql-with-schema](https://huggingface.co/juierror/flan-t5-text2sql-with-schema).
|
12 |
+
|
13 |
+
It supports the '<' sign and can handle multiple tables.
|
14 |
+
|
15 |
+
# How to use
|
16 |
+
```python
|
17 |
+
from typing import List
|
18 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
19 |
+
|
20 |
+
tokenizer = AutoTokenizer.from_pretrained("juierror/flan-t5-text2sql-with-schema-v2")
|
21 |
+
model = AutoModelForSeq2SeqLM.from_pretrained("juierror/flan-t5-text2sql-with-schema-v2")
|
22 |
+
|
23 |
+
def get_prompt(tables, question):
|
24 |
+
prompt = f"""convert question and table into SQL query. tables: {tables}. question: {question}"""
|
25 |
+
return prompt
|
26 |
+
|
27 |
+
def prepare_input(question: str, tables: Dict[str, List[str]]):
|
28 |
+
tables = [f"""{table_name}({",".join(tables[table_name])})""" for table_name in tables]
|
29 |
+
tables = ", ".join(tables)
|
30 |
+
prompt = get_prompt(tables, question)
|
31 |
+
input_ids = tokenizer(prompt, max_length=512, return_tensors="pt").input_ids
|
32 |
+
return input_ids
|
33 |
+
|
34 |
+
def inference(question: str, tables: Dict[str, List[str]]) -> str:
|
35 |
+
input_data = prepare_input(question=question, tables=tables)
|
36 |
+
input_data = input_data.to(model.device)
|
37 |
+
outputs = model.generate(inputs=input_data, num_beams=10, top_k=10, max_length=512)
|
38 |
+
result = tokenizer.decode(token_ids=outputs[0], skip_special_tokens=True)
|
39 |
+
return result
|
40 |
+
|
41 |
+
print(inference("how many people with name jui and age less than 25", {
|
42 |
+
"people_name": ["id", "name"],
|
43 |
+
"people_age": ["people_id", "age"]
|
44 |
+
}))
|
45 |
+
|
46 |
+
print(inference("what is id with name jui and age less than 25", {
|
47 |
+
"people_name": ["id", "name", "age"]
|
48 |
+
})))
|
49 |
+
```
|