import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
import torch

def load_model():
    model_name = "TheBloke/Llama-2-13B-chat-GGUF"  # Zmiana na publicznie dostępny model
    
    # Konfiguracja kwantyzacji 4-bitowej
    quantization_config = BitsAndBytesConfig(
        load_in_4bit=True,
        bnb_4bit_compute_dtype=torch.float16,
        bnb_4bit_quant_type="nf4",
        bnb_4bit_use_double_quant=True
    )
    
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForCausalLM.from_pretrained(
        model_name,
        device_map="auto",
        quantization_config=quantization_config,
        trust_remote_code=True
    )
    return model, tokenizer
def generate_response(prompt, max_length=100):
    try:
        inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
        outputs = model.generate(
            **inputs,
            max_new_tokens=max_length,
            temperature=0.7,
            top_p=0.9,
            repetition_penalty=1.2,
            do_sample=True
        )
        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
        return response
    except Exception as e:
        return f"Error: {str(e)}"

print("Ładowanie modelu...")
model, tokenizer = load_model()
print("Model załadowany!")

iface = gr.Interface(
    fn=generate_response,
    inputs=[
        gr.Textbox(label="Prompt", lines=5),
        gr.Slider(minimum=1, maximum=500, value=100, label="Max Length")
    ],
    outputs=gr.Textbox(label="Response", lines=5),
    title="Llama 2 Chat Bot",
    description="Bot RPG oparty na Llama 2"
)

iface.launch()