app.py

import torch
from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer, pipeline
import gradio as gr


peft_model_id = "Pr123/TinyLlama-EA-Chat"

# Load Model with PEFT adapter
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = AutoPeftModelForCausalLM.from_pretrained(
    peft_model_id,
    torch_dtype=torch.bfloat16
).to(device)

tokenizer = AutoTokenizer.from_pretrained(peft_model_id)
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_length=500)


def chat_with_tinyllm(prompt):
    instruction = "Answer the following question: if you don't know the answer, just say that you don't know; don't try to make up an answer."
    prompt_content = f"<s>[INST] <<SYS>>{instruction}<</SYS>>{prompt}[/INST]"
    result = pipe(prompt_content)
    result = result[0]['generated_text'].split('[/INST]')[-1]
    return result

def chat_interface():
    iface = gr.Interface(
        fn=chat_with_tinyllm,
        inputs=gr.Textbox(lines=2, placeholder="Type your question here..."),  
        outputs="text",
        title="Chat with TinyLlama",
        description="This is a simple chatbot powered by a fine-tuned model on Hugging Face. If it doesn't know the answer, it will say so.")
    return iface

iface = chat_interface()
iface.launch()