README.md

metadata

license: cc-by-sa-4.0
language:
  - ru
  - en
pipeline_tag: text-to-image
tags:
  - PyTorch
  - Transformers

Sunset Cities

This is the Malevich ruDALL-E model finetuned on anime screenshots of big cities at sunset.

installation

pip install rudalle

How to use

Basic implementation to get a list of image data objects.

from translate import Translator
from rudalle import get_rudalle_model, get_tokenizer, get_vae
from rudalle.pipelines import generate_images

model = get_rudalle_model('Malevich', pretrained=True, fp16=True, device='cuda')
model.load_state_dict(torch.load(CHECKPOINT_PATH))
vae = get_vae().to('cuda')
tokenizer = get_tokenizer()
input_text = Translator(to_lang='ru').translate('city at sunset')
images, _ = generate_images(
        text=input_text,
        tokenizer=tokenizer, dalle=model, vae=vae,
        images_num=1,
        top_k=2048,
        top_p=0.95,
        temperature=1.0
    )

the Malevich model only recognizes input in Russian. If you're going to paste Cyrillic directly into the code rather than filter an English prompt through the translate API, you will need to put this at the top of the file:

#!/usr/bin/env python3
# -*- coding: utf-8 -*-