AI說書 - 從0開始 - 532 | Stable Diffusion 之 Keras 實現 (流程串接)

我想要一天分享一點「LLM從底層堆疊的技術」，並且每篇文章長度控制在三分鐘以內，讓大家不會壓力太大，但是又能夠每天成長一點。

Keras 實現 Stable Diffusion 的必要安裝包為：

try:
	import tensorflow as tf
	print(tf.__version__)
except:
	!pip install tensorflow
	import tensorflow as tf
	print(tf.__version__)
	
!pip install keras_cv --upgrade --quiet
!pip install keras_core --upgrade --quiet

import time
import keras_cv
from tensorflow import keras
import matplotlib.pyplot as plt

model = keras_cv.models.StableDiffusion(img_width = 512, img_height = 512)

整體流程的描述為：The model functions as we saw in the previous sections. The model will run the text encoder, then gradually “denoise” a 64 x 64 latent image patch with the diffusion model. Finally, it will transform the “denoised” 64 x 64 patch into a higher resolution (sometimes referred to as “super-resolution”) 512 x 512 image