min-dalle-test/min_dalle/image_from_text.py

70 lines
2.2 KiB
Python
Raw Normal View History

2022-06-27 15:57:56 +00:00
import os
import json
import numpy
import torch
from PIL import Image
from typing import Tuple, List
2022-06-27 16:43:47 +00:00
from .text_tokenizer import TextTokenizer
from .models.vqgan_detokenizer import VQGanDetokenizer
from .load_params import load_vqgan_torch_params
2022-06-27 15:57:56 +00:00
def load_dalle_bart_metadata(path: str) -> Tuple[dict, dict, List[str]]:
print("loading model")
for f in ['config.json', 'flax_model.msgpack', 'vocab.json', 'merges.txt']:
assert(os.path.exists(os.path.join(path, f)))
with open(path + '/config.json', 'r') as f:
config = json.load(f)
with open(path + '/vocab.json') as f:
vocab = json.load(f)
with open(path + '/merges.txt') as f:
merges = f.read().split("\n")[1:-1]
return config, vocab, merges
def ascii_from_image(image: Image.Image, size: int) -> str:
rgb_pixels = image.resize((size, int(0.55 * size))).convert('L').getdata()
chars = list('.,;/IOX')
chars = [chars[i * len(chars) // 256] for i in rgb_pixels]
chars = [chars[i * size: (i + 1) * size] for i in range(size // 2)]
return '\n'.join(''.join(row) for row in chars)
def save_image(image: numpy.ndarray, path: str) -> Image.Image:
if os.path.isdir(path):
path = os.path.join(path, 'generated.png')
elif not path.endswith('.png'):
path += '.png'
print("saving image to", path)
image: Image.Image = Image.fromarray(numpy.asarray(image))
image.save(path)
return image
def tokenize(
text: str,
config: dict,
vocab: dict,
merges: List[str]
) -> numpy.ndarray:
print("tokenizing text")
tokens = TextTokenizer(vocab, merges)(text)
print("text tokens", tokens)
text_tokens = numpy.ones((2, config['max_text_length']), dtype=numpy.int32)
text_tokens[0, :len(tokens)] = tokens
text_tokens[1, :2] = [tokens[0], tokens[-1]]
return text_tokens
def detokenize_torch(
image_tokens: numpy.ndarray,
model_path: str
) -> numpy.ndarray:
print("detokenizing image")
params = load_vqgan_torch_params(model_path)
detokenizer = VQGanDetokenizer()
detokenizer.load_state_dict(params)
image_tokens = torch.tensor(image_tokens).to(torch.long)
image = detokenizer.forward(image_tokens).to(torch.uint8)
return image.detach().numpy()