app.py

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

from peft import PeftModel
import torch
import os 

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""

# Set your model and adapter paths
API_KEY = os.environ.get("llama_ACCESS_TOKEN")
BASE_MODEL = "meta-llama/Meta-Llama-3-8B"
PEFT_ADAPTER = "asdc/Llama-3-8B-multilingual-temporal-expression-normalization"

tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, token=API_KEY)
base_model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    torch_dtype=torch.float16,
    device_map="auto",
    token=API_KEY
)

nf4_config = BitsAndBytesConfig(
   load_in_4bit=True,
   bnb_4bit_quant_type="nf4",
   bnb_4bit_use_double_quant=True,
   bnb_4bit_compute_dtype=torch.bfloat16
)

model = PeftModel.from_pretrained(base_model, PEFT_ADAPTER, token=API_KEY, quantization_config=nf4_config)

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


def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    prompt = system_message + "\n"
    for user, assistant in history:
        if user:
            prompt += f"User: {user}\n"
        if assistant:
            prompt += f"Assistant: {assistant}\n"
    prompt += f"User: {message}\nAssistant:"

    outputs = pipe(
        prompt,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id,
    )
    response = outputs[0]["generated_text"][len(prompt):]
    yield response


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)


if __name__ == "__main__":
    demo.launch()