Qwen2.5-0.5B Text-to-SQL Model

This model is a fine-tuned version of Qwen2.5-0.5B for the text-to-SQL task on the b-mc2/sql-create-context dataset. It converts natural language questions + database schemas into SQL queries.

🔧 How to Use

1. Install dependencies:

   pip install transformers torch
   

2. Load the model:

    from transformers import AutoModelForCausalLM, AutoTokenizer
    import torch
    
    model_name = "iamkushagratomar/Qwen2.5-0.5B-text2sql"
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForCausalLM.from_pretrained(model_name).to("cuda")
   

3. Use the model (with your question + schema)::

   def generate_sql(question, schema):
     # Format input as: "System: [message]\n\nUser: [question + schema]"
     system_message = "You are a SQL expert. Convert natural language questions into SQL queries."
     user_message = f"### Question: \n{question}\n\n### Schema: \n{schema}"
     
     full_input = f"System: {system_message}\n\nUser: {user_message}"
     
     inputs = tokenizer(full_input, return_tensors="pt").to(model.device)
     outputs = model.generate(
         input_ids=inputs.input_ids,
         max_new_tokens=100,
         temperature=0.2,
         do_sample=True,
         use_cache=True
     )
     return tokenizer.decode(outputs[0], skip_special_tokens=True)
   
   # Example usage
   question = "Find the total revenue from orders placed in the last 30 days."
   schema = "CREATE TABLE orders (order_id INT, order_date DATE, order_total DECIMAL, customer_id INT);"
   sql_query = generate_sql(question, schema)
   print(sql_query)
   

Uploaded finetuned model

• Developed by: iamkushagratomar
• License: apache-2.0
• Finetuned from model : unsloth/Qwen2.5-0.5B-Instruct-bnb-4bit

This qwen2 model was trained 2x faster with Unsloth and Huggingface's TRL library.