Spaces:
Sleeping
Sleeping
| import torch | |
| from transformers import AutoTokenizer, AutoModelForCausalLM | |
| from peft import PeftConfig, PeftModel | |
| import gradio as gr | |
| import spaces | |
| MODEL_PATH = "Ozaii/zephyr-bae" | |
| BASE_MODEL = "unsloth/llama-3-8b-bnb-4bit" | |
| max_seq_length = 2048 | |
| print("Zephyr is getting ready to charm! 🌟") | |
| model = None | |
| tokenizer = None | |
| def load_model(): | |
| global model, tokenizer | |
| if model is None: | |
| try: | |
| peft_config = PeftConfig.from_pretrained(MODEL_PATH) | |
| base_model = AutoModelForCausalLM.from_pretrained( | |
| BASE_MODEL, | |
| torch_dtype=torch.float16, | |
| device_map="auto", | |
| trust_remote_code=True | |
| ) | |
| model = PeftModel.from_pretrained(base_model, MODEL_PATH) | |
| tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL) | |
| tokenizer.pad_token = tokenizer.eos_token | |
| print("Zephyr loaded successfully! Time to charm!") | |
| except Exception as e: | |
| print(f"Oops! Zephyr seems to be playing hide and seek. Error: {str(e)}") | |
| raise | |
| return model, tokenizer | |
| def generate_response(prompt, max_new_tokens=128): | |
| model, tokenizer = load_model() | |
| inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=2048).to(model.device) | |
| outputs = model.generate( | |
| **inputs, | |
| max_new_tokens=max_new_tokens, | |
| temperature=0.7, | |
| top_p=0.9, | |
| repetition_penalty=1.2, | |
| do_sample=True | |
| ) | |
| return tokenizer.decode(outputs[0], skip_special_tokens=True) | |
| def chat_with_zephyr(message, history): | |
| conversation_history = history[-3:] # Limit to last 3 exchanges | |
| full_prompt = "\n".join([f"Human: {h[0]}\nZephyr: {h[1]}" for h in conversation_history]) | |
| full_prompt += f"\nHuman: {message}\nZephyr:" | |
| response = generate_response(full_prompt) | |
| # Extract Zephyr's response | |
| zephyr_response = response.split("Zephyr:")[-1].strip() | |
| # Append the new exchange to the history | |
| history.append((message, zephyr_response)) | |
| return history | |
| css = """ | |
| body { | |
| background-color: #1a1a2e; | |
| color: #e0e0ff; | |
| } | |
| .gradio-container { | |
| background-color: #1a1a2e; | |
| } | |
| """ | |
| with gr.Blocks(css=css) as iface: | |
| gr.Markdown("# Chat with Zephyr: Your AI Boyfriend 💘") | |
| chatbot = gr.Chatbot(height=500) | |
| msg = gr.Textbox(placeholder="Tell Zephyr what's on your mind...", label="Your message") | |
| clear = gr.Button("Clear Chat") | |
| def respond(message, chat_history): | |
| bot_message = chat_with_zephyr(message, chat_history) | |
| chat_history.append((message, bot_message[-1][1])) | |
| return "", chat_history | |
| msg.submit(respond, [msg, chatbot], [msg, chatbot]) | |
| clear.click(lambda: None, None, chatbot, queue=False) | |
| gr.Markdown(""" | |
| ## Welcome to Zephyr, Your AI Boyfriend! | |
| Zephyr is here to charm you with his wit, humor, and cosmic energy. Feel free to flirt, ask for advice, or just chat about anything under the stars! | |
| **Some conversation starters:** | |
| - "Hey Zephyr, how's the cosmic energy today?" | |
| - "What's your idea of a perfect date in the digital realm?" | |
| - "Tell me something that would make me fall for you even more!" | |
| Remember, Zephyr is an AI and this is for fun and entertainment. Enjoy your chat! 💖 | |
| """) | |
| if __name__ == "__main__": | |
| iface.launch() |