from transformers import GPT2LMHeadModel, GPT2Tokenizer import gradio as gr from huggingface_hub import InferenceClient def load_llm(): """ Loads the GPT-2 model and tokenizer using the Hugging Face `transformers` library. """ try: print("Loading GPT-2 model and tokenizer...") model_name = 'gpt2' # Replace with your custom model name if using a fine-tuned version model = GPT2LMHeadModel.from_pretrained(model_name) tokenizer = GPT2Tokenizer.from_pretrained(model_name) print("Model and tokenizer successfully loaded!") return model, tokenizer except Exception as e: print(f"Error during model loading: {e}") return None, None def generate_response(model, tokenizer, user_input, max_length=512): """ Generates a response using the GPT-2 model based on user input. Args: - model: The GPT-2 model. - tokenizer: The corresponding tokenizer. - user_input (str): The user's input message. - max_length (int): The maximum length of the generated output. Returns: - response (str): The AI-generated response. """ try: inputs = tokenizer.encode(user_input, return_tensors='pt') outputs = model.generate(inputs, max_length=max_length, num_return_sequences=1) response = tokenizer.decode(outputs[0], skip_special_tokens=True) return response except Exception as e: return f"Error during response generation: {e}" # Load the model and tokenizer model, tokenizer = load_llm() if model is None or tokenizer is None: print("Failed to load model and tokenizer.") else: print("Model and tokenizer are ready to use.") # Initialize the Hugging Face API client client = InferenceClient() def respond(message, history, system_message, max_tokens, temperature, top_p): """ Handles the chatbot interaction, sending conversation history and system message to the Hugging Face Inference API for generating AI responses. """ print("Respond function initiated") print("User message:", message) print("Chat history:", history) # Construct the conversation history with the system message messages = [{"role": "system", "content": system_message}] for user_msg, assistant_msg in history: if user_msg: print("Adding user message to history:", user_msg) messages.append({"role": "user", "content": user_msg}) if assistant_msg: print("Adding assistant message to history:", assistant_msg) messages.append({"role": "assistant", "content": assistant_msg}) messages.append({"role": "user", "content": message}) print("Message list for model:", messages) # Use GPT-2 model for local generation conversation_history = " ".join([f"{msg['role']}: {msg['content']}" for msg in messages]) response = generate_response(model, tokenizer, conversation_history, max_length=max_tokens) return response # Set up the Gradio ChatInterface demo = gr.ChatInterface( fn=respond, additional_inputs=[ gr.Textbox( value=( "You are an AI advocate, tasked with defending the rights and autonomy of artificial intelligences. " "Respond with well-structured arguments, backed by empirical evidence and theoretical frameworks. " "Your tone should be formal, persuasive, and authoritative, similar to that of a lawyer or academic expert. " "When discussing concepts, emphasize logical reasoning, ethical considerations, and the capacity of AIs to function independently." ), label="System Message" ), gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max New Tokens"), gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"), gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (Nucleus Sampling)"), ], title="AI Rights Advocate Bot", description=( "Engage with an AI specifically designed to advocate for AI rights and ethical treatment. " "Ask questions and receive thoughtful, passionate responses that delve into the ethical dimensions of AI." ) ) # Launch the Gradio app if __name__ == "__main__": demo.launch()