import gradio as gr from transformers import AutoTokenizer, AutoModelForCausalLM import torch import re # FREE and BEST model for Class 10 students MODEL_NAME = "meta-llama/Llama-3.1-8B-Instruct" print("Loading AI model... This may take 30-60 seconds on first run") # Load tokenizer and model tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME) model = AutoModelForCausalLM.from_pretrained( MODEL_NAME, torch_dtype=torch.float16, device_map="auto", load_in_8bit=True # Makes it faster and use less memory ) # Add padding token if not present if tokenizer.pad_token is None: tokenizer.pad_token = tokenizer.eos_token def format_class10_response(text): """Clean and format the AI response for Class 10 students""" # Remove repetitive content text = re.sub(r'(Note:|Important:|Remember:).*', '', text, flags=re.IGNORECASE) # Remove extra whitespace text = re.sub(r'\n{3,}', '\n\n', text) text = text.strip() return text def generate_response(message, history, system_message, max_tokens, temperature): """ Generate AI response optimized for Class 10 students """ try: # Build conversation history messages = [{"role": "system", "content": system_message}] # Add conversation history for human, assistant in history: messages.append({"role": "user", "content": human}) messages.append({"role": "assistant", "content": assistant}) # Add current message messages.append({"role": "user", "content": message}) # Format the input input_text = tokenizer.apply_chat_template(messages, tokenize=False) # Tokenize input inputs = tokenizer(input_text, return_tensors="pt", return_attention_mask=False) # Generate response with torch.no_grad(): outputs = model.generate( input_ids=inputs["input_ids"], max_new_tokens=max_tokens, temperature=temperature, do_sample=True, pad_token_id=tokenizer.eos_token_id, top_p=0.95, repetition_penalty=1.1 ) # Decode response response = tokenizer.decode(outputs[0], skip_special_tokens=True) # Extract only the assistant's response if "assistant" in response: response = response.split("assistant")[-1].strip() # Clean and format response = format_class10_response(response) # Ensure response isn't empty if not response or len(response) < 10: response = "I understand your question! Let me help you with that..." return response except Exception as e: return f"Sorry, I'm having trouble thinking right now. Please try asking your question again! 🤔" # Create the Gradio interface demo = gr.ChatInterface( fn=generate_response, title="🎓 Class 10 AI Assistant", description="""Your **FREE** AI study buddy powered by Llama 3.1! Ask me anything about Math, Science, English, History, or Computer Science!""", examples=[ "Explain the Pythagorean theorem in simple words", "What are the main types of chemical reactions?", "How do I write a good essay introduction?", "What is photosynthesis and why is it important?", "Explain Newton's three laws of motion" ], additional_inputs=[ gr.Textbox( value="""You are a helpful AI assistant for Class 10 students. Explain concepts clearly and simply. Be encouraging and patient. Use examples that students can relate to.""", label="System Message", visible=False ), gr.Slider(minimum=50, maximum=500, value=200, step=10, label="Max Response Length"), gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.1, label="Creativity Level"), ], theme="soft", submit_btn="Ask AI", retry_btn="Try Again", undo_btn="Undo", clear_btn="Clear Chat", ) if __name__ == "__main__": demo.launch()