Diffusion_from_scratch

Introduction

从0到1手写基于mnist手写数字数据集的stable diffusion模型复现

引入cross-attention来将cls-embedding放入unet中, 来实现可控图像生成, 即stable diffusion

Name		Name	Last commit message	Last commit date
Latest commit History 20 Commits
assets		assets
notebook		notebook
.gitignore		.gitignore
README.md		README.md
config.py		config.py
conv_block.py		conv_block.py
cross_attention.py		cross_attention.py
dataset.py		dataset.py
denoise.py		denoise.py
diffusion.py		diffusion.py
time_position_emb.py		time_position_emb.py
train.py		train.py
unet.py		unet.py