app.py

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

tokenizer = AutoTokenizer.from_pretrained(
  'kakaobrain/kogpt', revision='KoGPT6B-ryan1.5b-float16',  # or float32 version: revision=KoGPT6B-ryan1.5b
  bos_token='[BOS]', eos_token='[EOS]', unk_token='[UNK]', pad_token='[PAD]', mask_token='[MASK]'
)

model = AutoModelForCausalLM.from_pretrained(
  'kakaobrain/kogpt', revision='KoGPT6B-ryan1.5b-float16',  # or float32 version: revision=KoGPT6B-ryan1.5b
  pad_token_id=tokenizer.eos_token_id,
  torch_dtype='auto', low_cpu_mem_usage=True
).to(device='cpu', non_blocking=True)
_ = model.eval()

title = "KoGPT"
description = "Gradio demo for KoGPT(Korean Generative Pre-trained Transformer). To use it, simply add your text, or click one of the examples to load them. Read more at the links below."
article = "<p style='text-align: center'><a href='https://github.com/kakaobrain/kogpt' target='_blank'>KoGPT: KakaoBrain Korean(hangul) Generative Pre-trained Transformer</a> | <a href='https://huggingface.co/kakaobrain/kogpt' target='_blank'>Huggingface Model</a></p>"
examples=[['인간처럼 생각하고, 행동하는 \'지능\'을 통해 인류가 이제까지 풀지 못했던']]
def greet(text):
    prompt = text
    with torch.no_grad():
      tokens = tokenizer.encode(prompt, return_tensors='pt').to(device='cpu', non_blocking=True)
      gen_tokens = model.generate(tokens, do_sample=True, temperature=0.8, max_length=64)
      generated = tokenizer.batch_decode(gen_tokens)[0]

    print(f"generated {generated}")
      
    return generated
    
iface = gr.Interface(fn=greet, inputs="text", outputs="text", title=title, description=description, article=article, examples=examples,enable_queue=True)
iface.launch()