initial commit

README.md

@@ -5,8 +5,9 @@ colorFrom: red
 colorTo: red
 sdk: gradio
 sdk_version: 5.11.0
-app_file: app.py
+app_file: gpt_neo.db.py
 pinned: false
+python: 3.9
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

gpt_neo_db.py

@@ -0,0 +1,61 @@
+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-neox-20b"  # 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):
+    """
+    Generate an SQL query based on the question and context.
+
+    Args:
+        context (str): Description of the database schema or table relationships.
+        question (str): User's natural language query.
+
+    Returns:
+        str: Generated SQL query.
+    """
+    # Prepare the prompt
+    prompt = f"""
+    Context: {context}
+
+    Question: {question}
+
+    Write an SQL query to address the question based on the context.
+    Query:
+    """
+    # Tokenize input
+    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=1024).to("cuda" if torch.cuda.is_available() else "cpu")
+
+    # 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
+print("Ask a question about the database schema.")
+while True:
+    user_question = input("Question: ")
+    if user_question.lower() in ["exit", "quit"]:
+        print("Exiting...")
+        break
+
+    # Generate SQL query
+    sql_query = generate_sql_query(schema_description, user_question)
+    print(f"Generated SQL Query:\n{sql_query}\n")

requirements.txt

@@ -0,0 +1,3 @@
+transformers
+torch
+json