import time
import gradio as gr
from ctransformers import AutoModelForCausalLM

# Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.
llm = AutoModelForCausalLM.from_pretrained("TheBloke/WizardLM-7B-uncensored-GGUF", model_file="WizardLM-7B-uncensored.Q4_K_M.gguf", model_type="llama", stream=True)

def generate_response(message):
    tokens = llm(message)
    response_sentence = ' '.join(tokens)
    yield response_sentence
    

def chatbot(message, history):
    response_generator = generate_response(message)
    for response in response_generator:
        time.sleep(0.3)
        # Optional delay for a natural chat feel
        
        yield response 

iface = gr.ChatInterface(chatbot)
iface.launch()