File size: 2,036 Bytes
4c735b8
 
 
 
 
 
 
 
 
 
 
9c8236d
c1abd5c
9c8236d
 
4c735b8
9c8236d
 
83ce0d2
 
9c8236d
 
 
 
 
 
83ce0d2
9c8236d
 
 
 
 
 
 
 
 
c1abd5c
 
 
 
9c8236d
 
 
 
 
 
 
 
4c735b8
 
 
 
 
9c8236d
c8e1797
9c8236d
4c735b8
9c8236d
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
import json
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Define the schema for the database
db_schema = {
    "products": ["product_id", "name", "price", "description", "type"],
    "orders": ["order_id", "product_id", "quantity", "order_date"],
    "customers": ["customer_id", "name", "email", "phone_number"]
}

# Load the model and tokenizer
model_name = "EleutherAI/gpt-neo-2.7B"  # You can also use "Llama-2-7b" or another model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", torch_dtype=torch.float16)

def generate_sql_query(context, question):
    """
    This is the description of the database which is given to you, a user can ask
    anything related to this database

    Args:
        context (str): Description of the database schema or table relationships.
        question (str): User's natural language query.

    Returns:
        str: An answer to the question.
    """
    # Prepare the prompt
    prompt = f"""
    Context: {context}

    Question: {question}
    Query:
    """
    # Tokenize input
    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=512).to("cuda" if torch.cuda.is_available() else "cpu")

    print("Prompt Sent to Model:")
    print(prompt)

    # Generate SQL query
    output = model.generate(inputs.input_ids, max_length=512, num_beams=5, early_stopping=True)
    query = tokenizer.decode(output[0], skip_special_tokens=True)

    # Extract query from the output
    sql_query = query.split("Query:")[-1].strip()
    return sql_query

# Schema as a context for the model
schema_description = json.dumps(db_schema, indent=4)

# Example interactive questions
questions = [
    "describe the product table for me, what kind of data it is storing and all"
]

for user_question in questions:
    print(f"Question: {user_question}")
    sql_query = generate_sql_query(schema_description, user_question)
    print(f"Generated SQL Query:\n{sql_query}\n")