Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,24 +1,52 @@
|
|
1 |
import gradio as gr
|
|
|
2 |
from datasets import load_dataset
|
3 |
|
4 |
# Load the WikiSQL dataset
|
5 |
wikisql_dataset = load_dataset("wikisql", split='train[:100]') # Load a subset of the dataset
|
6 |
|
7 |
-
#
|
8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
def generate_sql_from_user_input(query):
|
11 |
-
#
|
12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
return sql_query
|
14 |
|
15 |
# Create a Gradio interface
|
16 |
interface = gr.Interface(
|
17 |
fn=generate_sql_from_user_input,
|
18 |
inputs=gr.Textbox(label="Enter your natural language query"),
|
19 |
-
outputs=gr.Textbox(label="SQL Query
|
20 |
-
title="NL to SQL using WikiSQL Dataset",
|
21 |
-
description="This
|
22 |
)
|
23 |
|
24 |
# Launch the app
|
|
|
1 |
import gradio as gr
|
2 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
3 |
from datasets import load_dataset
|
4 |
|
5 |
# Load the WikiSQL dataset
|
6 |
wikisql_dataset = load_dataset("wikisql", split='train[:100]') # Load a subset of the dataset
|
7 |
|
8 |
+
# Extract schema information from the dataset
|
9 |
+
table_names = set()
|
10 |
+
column_names = set()
|
11 |
+
for item in wikisql_dataset:
|
12 |
+
table_names.add(item['table']['name'])
|
13 |
+
for column in item['table']['header']:
|
14 |
+
column_names.add(column)
|
15 |
+
|
16 |
+
# Load tokenizer and model
|
17 |
+
tokenizer = AutoTokenizer.from_pretrained("mrm8488/t5-base-finetuned-wikiSQL")
|
18 |
+
model = AutoModelForSeq2SeqLM.from_pretrained("mrm8488/t5-base-finetuned-wikiSQL")
|
19 |
+
|
20 |
+
def post_process_sql_query(sql_query):
|
21 |
+
# Modify the SQL query to match the dataset's schema
|
22 |
+
# This is just an example and might need to be adapted based on the dataset and model output
|
23 |
+
for table_name in table_names:
|
24 |
+
if "TABLE" in sql_query:
|
25 |
+
sql_query = sql_query.replace("TABLE", table_name)
|
26 |
+
break # Assuming only one table is referenced in the query
|
27 |
+
for column_name in column_names:
|
28 |
+
if "COLUMN" in sql_query:
|
29 |
+
sql_query = sql_query.replace("COLUMN", column_name, 1)
|
30 |
+
return sql_query
|
31 |
|
32 |
def generate_sql_from_user_input(query):
|
33 |
+
# Generate SQL for the user's query
|
34 |
+
input_text = "translate English to SQL: " + query
|
35 |
+
inputs = tokenizer(input_text, return_tensors="pt", padding=True)
|
36 |
+
outputs = model.generate(**inputs, max_length=512)
|
37 |
+
sql_query = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
38 |
+
|
39 |
+
# Post-process the SQL query to match the dataset's schema
|
40 |
+
sql_query = post_process_sql_query(sql_query)
|
41 |
return sql_query
|
42 |
|
43 |
# Create a Gradio interface
|
44 |
interface = gr.Interface(
|
45 |
fn=generate_sql_from_user_input,
|
46 |
inputs=gr.Textbox(label="Enter your natural language query"),
|
47 |
+
outputs=gr.Textbox(label="Generated SQL Query"),
|
48 |
+
title="NL to SQL with T5 using WikiSQL Dataset",
|
49 |
+
description="This model generates an SQL query for your natural language input based on the WikiSQL dataset."
|
50 |
)
|
51 |
|
52 |
# Launch the app
|