Manoj Kumar
commited on
Commit
·
9c8236d
1
Parent(s):
39179ce
updated question structure
Browse files
app.py
CHANGED
@@ -9,21 +9,52 @@ db_schema = {
|
|
9 |
"customers": ["customer_id", "name", "email", "phone_number"]
|
10 |
}
|
11 |
|
|
|
|
|
|
|
|
|
12 |
|
13 |
-
def
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
# Schema as a context for the model
|
17 |
schema_description = json.dumps(db_schema, indent=4)
|
18 |
|
19 |
# Example interactive questions
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
print("Exiting...")
|
25 |
-
break
|
26 |
|
27 |
-
|
28 |
-
|
29 |
-
|
|
|
|
9 |
"customers": ["customer_id", "name", "email", "phone_number"]
|
10 |
}
|
11 |
|
12 |
+
# Load the model and tokenizer
|
13 |
+
model_name = "EleutherAI/gpt-neox-20b" # You can also use "Llama-2-7b" or another model
|
14 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
15 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", torch_dtype=torch.float16)
|
16 |
|
17 |
+
def generate_sql_query(context, question):
|
18 |
+
"""
|
19 |
+
Generate an SQL query based on the question and context.
|
20 |
+
|
21 |
+
Args:
|
22 |
+
context (str): Description of the database schema or table relationships.
|
23 |
+
question (str): User's natural language query.
|
24 |
+
|
25 |
+
Returns:
|
26 |
+
str: Generated SQL query.
|
27 |
+
"""
|
28 |
+
# Prepare the prompt
|
29 |
+
prompt = f"""
|
30 |
+
Context: {context}
|
31 |
+
|
32 |
+
Question: {question}
|
33 |
+
|
34 |
+
Write an SQL query to address the question based on the context.
|
35 |
+
Query:
|
36 |
+
"""
|
37 |
+
# Tokenize input
|
38 |
+
inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=1024).to("cuda" if torch.cuda.is_available() else "cpu")
|
39 |
+
|
40 |
+
# Generate SQL query
|
41 |
+
output = model.generate(inputs.input_ids, max_length=512, num_beams=5, early_stopping=True)
|
42 |
+
query = tokenizer.decode(output[0], skip_special_tokens=True)
|
43 |
+
|
44 |
+
# Extract query from the output
|
45 |
+
sql_query = query.split("Query:")[-1].strip()
|
46 |
+
return sql_query
|
47 |
|
48 |
# Schema as a context for the model
|
49 |
schema_description = json.dumps(db_schema, indent=4)
|
50 |
|
51 |
# Example interactive questions
|
52 |
+
questions = [
|
53 |
+
"Show all products that cost more than $50.",
|
54 |
+
"List all customers who ordered a specific product.",
|
55 |
+
]
|
|
|
|
|
56 |
|
57 |
+
for user_question in questions:
|
58 |
+
print(f"Question: {user_question}")
|
59 |
+
sql_query = generate_sql_query(schema_description, user_question)
|
60 |
+
print(f"Generated SQL Query:\n{sql_query}\n")
|