app.py

# app.py

import gradio as gr
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("trained_model", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("trained_model")

pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

def chatbot(instruction):
    prompt = f"### Instruction:\n{instruction}\n\n### Response:\n"
    response = pipe(prompt, max_new_tokens=100)[0]['generated_text']
    return response[len(prompt):].strip()

gr.Interface(
    fn=chatbot,
    inputs="text",
    outputs="text",
    title="TinyLlama QLoRA Support Bot"
).launch()