Spaces:
Sleeping
Sleeping
| import json | |
| import gradio as gr | |
| from huggingface_hub import InferenceClient | |
| # Function to call the Llama 3.1 8B model through Hugging Face API | |
| def call_llama_model(user_query): | |
| # Initialize the inference client | |
| client = InferenceClient("meta-llama/Meta-Llama-3.1-8B-Instruct") | |
| # Define the addition function schema | |
| function_schema = { | |
| "name": "add_numbers", | |
| "description": "Add two numbers together", | |
| "parameters": { | |
| "type": "object", | |
| "properties": { | |
| "num1": { | |
| "type": "number", | |
| "description": "First number to add" | |
| }, | |
| "num2": { | |
| "type": "number", | |
| "description": "Second number to add" | |
| } | |
| }, | |
| "required": ["num1", "num2"] | |
| } | |
| } | |
| # Create the system prompt with function definition | |
| system_prompt = f"""You have access to the following function: | |
| {json.dumps(function_schema, indent=2)} | |
| Your task is to extract two numbers from the user's query and call the add_numbers function. | |
| Format your response as JSON with the function name and parameters. | |
| Only respond with valid JSON containing the function call, nothing else. | |
| """ | |
| # Call the model | |
| response = client.text_generation( | |
| prompt=f"<|system|>\n{system_prompt}\n<|user|>\n{user_query}\n<|assistant|>", | |
| max_new_tokens=256, | |
| temperature=0.1, | |
| return_full_text=False | |
| ) | |
| return response | |
| # Function to parse the model response and calculate the result | |
| def process_addition(query): | |
| try: | |
| # Get model response | |
| model_response = call_llama_model(query) | |
| # Try to parse the JSON response | |
| try: | |
| # Find the JSON part in the response (it might have additional text) | |
| json_start = model_response.find('{') | |
| json_end = model_response.rfind('}') + 1 | |
| if json_start >= 0 and json_end > json_start: | |
| json_str = model_response[json_start:json_end] | |
| response_data = json.loads(json_str) | |
| else: | |
| return f"Error: No valid JSON found in response: {model_response}" | |
| # Check if it has a function call | |
| if "function_call" in response_data: | |
| function_name = response_data["function_call"]["name"] | |
| params = response_data["function_call"]["parameters"] | |
| if function_name == "add_numbers": | |
| num1 = params["num1"] | |
| num2 = params["num2"] | |
| result = num1 + num2 | |
| # Return a formatted response | |
| return f""" | |
| Model parsed your query as: | |
| - First number: {num1} | |
| - Second number: {num2} | |
| Function called: {function_name} | |
| Result: {result} | |
| """ | |
| else: | |
| return f"Unknown function: {function_name}" | |
| else: | |
| return f"No function call found in response: {response_data}" | |
| except json.JSONDecodeError as e: | |
| return f"Error parsing JSON: {str(e)}\nRaw response: {model_response}" | |
| except Exception as e: | |
| return f"Error: {str(e)}" | |
| # Create Gradio interface | |
| def create_demo(): | |
| with gr.Blocks() as demo: | |
| gr.Markdown("# Llama 3.1 Function Calling Demo: Addition") | |
| gr.Markdown("Enter a query asking to add two numbers (e.g., 'Add 25 and 17' or 'What's 42 plus 58?')") | |
| with gr.Row(): | |
| with gr.Column(): | |
| query_input = gr.Textbox( | |
| label="Your Query", | |
| placeholder="Add 25 and 17" | |
| ) | |
| submit_btn = gr.Button("Calculate") | |
| with gr.Column(): | |
| output = gr.Textbox(label="Result") | |
| submit_btn.click( | |
| fn=process_addition, | |
| inputs=query_input, | |
| outputs=output | |
| ) | |
| gr.Examples( | |
| examples=[ | |
| "Add 25 and 17", | |
| "What is 42 plus 58?", | |
| "Can you sum 123 and 456?", | |
| "I need to add 7.5 and 2.25", | |
| "What's the total of 1000 and 2000?" | |
| ], | |
| inputs=query_input | |
| ) | |
| return demo | |
| # Create and launch the demo | |
| demo = create_demo() | |
| if __name__ == "__main__": | |
| demo.launch() |