from transformers import AutoModelForCausalLM, AutoTokenizer
import gradio as gr

# Load the model and tokenizer
model_name = "defog/llama-3-sqlcoder-8b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Define the function for querying the model
def generate_sql(prompt):
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs, max_length=150)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Create a Gradio interface
iface = gr.Interface(fn=generate_sql, inputs="text", outputs="text")
iface.launch()