app.py

import gradio as gr
import spaces
import torch

import transformers
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "meta-llama/Meta-Llama-3-8B-Instruct"

pipeline = transformers.pipeline(
    "text-generation",
    model=model_name,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device="cpu",
)

@spaces.GPU
def chat_function(message, history, system_prompt,max_new_tokens,temperature):
    messages = [
        {"role": "system", "content": system_prompt},
        {"role": "user", "content": message},
    ]
    prompt = pipeline.tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
    )
    terminators = [
        pipeline.tokenizer.eos_token_id,
        pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
    ]
    temp = temperature + 0.1
    outputs = pipeline(
        prompt,
        max_new_tokens=max_new_tokens,
        eos_token_id=terminators,
        do_sample=True,
        temperature=temp,
        top_p=0.9,
    )
    return outputs[0]["generated_text"][len(prompt):]

gr.ChatInterface(
    chat_function,
    chatbot=gr.Chatbot(height=400),
    textbox=gr.Textbox(placeholder="Enter message here", container=False, scale=7),
    title="Meta-Llama-3-8B-Instruct",
    description="""
    Two Slow? Running on CPU 😒.
    """,
    theme="soft",
    additional_inputs=[
        gr.Textbox("You are helpful AI.", label="System Prompt"),
        gr.Slider(512, 4096, label="Max New Tokens"),
        gr.Slider(0, 1, label="Temperature")
    ]
).launch()