Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
[CVPR 2025] "DiC: Rethinking Conv3x3 Designs in Diffusion Models", a performant & speedy Conv3x3 diffusion model.
| Date | Stars |
|---|---|
| 2026-07-31 | 251 |
| 2026-08-04 | 251 |
| 2026-08-06 | 251 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# DiC: Rethinking Conv3x3 Designs in Diffusion Models
<p align="left">
<a href="https://arxiv.org/abs/2501.00603" alt="arXiv">
<img src="https://img.shields.io/badge/arXiv-2501.00603-b31b1b.svg?style=flat" /></a>
<a href="https://github.com/YuchuanTian/DiC/blob/main/imgs/poster_cvpr.pdf" alt="arXiv">
<img src="https://img.shields.io/badge/Poster-b31b1b.svg?style=flat" /></a>
</p>
**_ConvUNets have been overlooked... but they outperform Diffusion Transformers!_**
## News
**6/11/2025:** We have released the codes of DiC! 🔥🔥🔥 Weights, SiT, and REPA versions are coming very soon.
**3/3/2025:** Codes & Weights are at the final stage of inspection. We will have them released ASAP.
**2/27/2025:** DiC is accepted by CVPR 2025! 🎉🎉
🤔 In this work, we intend to build a diffusion model with Conv3x3 that is simple but efficient.
🔧 We re-design architectures & blocks of the model to tap the potential of Conv3x3 to the full.
🚀 The proposed DiC ConvUNets are more powerful than DiTs, and **much much faster**!
## Repo Outline
This repo is mostly based on the official repo of DiT. Weights, SiT and REPA versions will be opensourced very soon.
Torch model script: **dic_models.py**
## Preparation
Please run command ```pip install -r requirements.txt``` to install the supporting packages.
(Optional) Please download the VAE from this [link](https://huggingface.co/stabilityai/sd-vae-ft-ema). The VAE could be automatically downloaded as well.
## Training
Here we provide two ways to train a DiC model: 1. train on the original ImageNet dataset; 2. train on preprocessed VAE features (Recommended).
**Training Data Preparation**
Use the original ImageNet dataset + VAE encoder. Firstly, download ImageNet as follows:
```
imagenet/
├──train/
│ ├── n01440764
│ │ ├── n01440764_10026.JPEG
│ │ ├── n01440764_10027.JPEG
│ │ ├── ......
│ ├── ......
├──val/
│ ├── n01440764
│ │ ├── ILSVRC2012_val_00000293.JPEG
│ │ ├── ILSVRC2012_val_00002138.JPEG
│ │ ├── ......
│ ├── ......
```
Then run the following command:
```bash
torchrun --nnodes=1 --nproc_per_node=8 train.py --data-path={path to imagenet/train} --image-size=256 --model={model name} --epochs={iteration//5000} # fp32 Training
accelerate launch --mixed_precision fp16 train_accelerate.py --data-path {path to imagenet/train} --image-size=256 --model={model name} --epochs={iteration//5000} # fp16 Training
```
**Training Feature Preparation (RECOMMENDED)**
Following Fast-DiT, it is recommended to load VAE features directly for faster training. You don't need to download the enormous ImageNet dataset (> 100G); instead, a much smaller "VAE feature" dataset (~21G for ImageNet 256x256) is available here on [HuggingFace](https://huggingface.co/datasets/yuchuantian/imagenet_vae_256) and [MindScope](https://www.modelscope.cn/models/YuchuanTian/imagenet_vae_256/). Please do the following steps:
1. Download [imagenet_feature.tar](https://huggingface.co/datasets/yuchuantian/imagenet_vae_256/blob/main/imagenet_feature.tar)
2. Unzip the tar ball by running ```tar -xf imagenet_feature.tar```
```
imagenet_feature/
├── imagenet256_features/ # VAE features
└── imagenet256_labels/ # labels
```
3. Append parser ```--feature-path={path to imagenet_feature}``` to the training command.
## Inference
#### Weights
Coming soon. Please keep tuned!
#### Sampling
Run the following command for parallel sampling:
```bash
torch --nnodes=1 --nproc_per_node=8 sample_ddp.py --ckpt={path to checkpoint} --image-size=256 --model={model name} --cfg-scale={cfg scale}
```
## BibTex Formatted Citation
If you find this repo useful, please cite:
```
@article{tian2025dic,
author = {Yuchuan Tian and
Jing Han and
Chengcheng Wang and
Yuchen Liang and
Chao Xu and
Excerpt of 4,959 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:c21fa56e2e1ca06b, desc:diffusion model