app.py

import re
import os
import requests
import gradio as gr
from datasets import load_dataset
from PIL import Image
from io import BytesIO
import torch
from torch import autocast

from transformers import pipeline, set_seed
from diffusers import DiffusionPipeline, StableDiffusionPipeline


# Config
DEVICE = "cuda"

# GPT2
def get_gpt2_pipeline():
  generator = pipeline('text-generation', model='gpt2')
  set_seed(42)

  # generator("Hello world, I'm vizard,", max_length=50, num_return_sequences=3)
  
  return generator

# SD v1.4
def get_stable_diffusion_v14_pipeline():
  model_id = "CompVis/stable-diffusion-v1-4"
  pipe = StableDiffusionPipeline.from_pretrained(mode_id)
  # pipeline = StableDiffusionPipeline.from_pretrained(model_id, use_auth_token=True, revision="fp16", torch_dtype=torch.float16)
  pipe = pipe.to(DEVICE)
  torch.backends.cudnn.benchmark = True
  return pipe

# SD v1.5
def get_stable_diffusion_v15_pipeline():
  model_id = "runwayml/stable-diffusion-v1-5"
  pipe = DiffusionPipeline.from_pretrained(mode_id)
  pipe = pipe.to(DEVICE)
  return pipe

def get_image(url):
  response = requests.get(url)
  image = Image.open(BytesIO(response.content)).convert("RGB")
  resized_image = image.resize((768, 512))
  return resized_image

# main
def main():
  prompt = "Hello world, I'm vizard,"
  
  pipe = pipeline(task="image-classification", 
                model="microsoft/dit-base-finetuned-rvlcdip")
  gr.Interface.from_pipeline(pipe, 
                           title=title,
                           description=description,
                           examples=['coca_cola_advertisement.png', 'scientific_publication.png', 'letter.jpeg'],
                           article=article,
                           enable_queue=True,
                           ).launch()
  
  # pipe2 = get_stable_diffusion_v15_pipeline()
  # images = pipe2(prompt).images

main