Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/FoundationVision/VAR
[NeurIPS 2024 Oral][GPT beats diffusion🔥] [scaling laws in visual generation📈] Official impl. of "Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction". An *ultra-simple, user-friendly yet state-of-the-art* codebase for autoregressive image generation!
https://github.com/FoundationVision/VAR
auto-regressive-model autoregressive-models diffusion-models generative-ai generative-model gpt gpt-2 image-generation large-language-models neurips transformers vision-transformer
Last synced: 13 days ago
JSON representation
[NeurIPS 2024 Oral][GPT beats diffusion🔥] [scaling laws in visual generation📈] Official impl. of "Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction". An *ultra-simple, user-friendly yet state-of-the-art* codebase for autoregressive image generation!
- Host: GitHub
- URL: https://github.com/FoundationVision/VAR
- Owner: FoundationVision
- License: mit
- Created: 2024-04-01T16:53:18.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2024-10-06T16:08:28.000Z (about 1 month ago)
- Last Synced: 2024-10-29T15:27:17.935Z (19 days ago)
- Topics: auto-regressive-model, autoregressive-models, diffusion-models, generative-ai, generative-model, gpt, gpt-2, image-generation, large-language-models, neurips, transformers, vision-transformer
- Language: Python
- Homepage:
- Size: 58.6 KB
- Stars: 4,194
- Watchers: 116
- Forks: 309
- Open Issues: 30
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-multi-modal - https://github.com/FoundationVision/VAR
- awesome-multi-modal - https://github.com/FoundationVision/VAR
- AiTreasureBox - FoundationVision/VAR - 11-13_4249_5](https://img.shields.io/github/stars/FoundationVision/VAR.svg)|[GPT beats diffusion🔥] [scaling laws in visual generation📈] Official impl. of "Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction"| (Repos)
- StarryDivineSky - FoundationVision/VAR
- awesome-llm-projects - VAR - style models beyond diffusion & Scaling laws observed. (Projects / 🌄 Image)
README
# VAR: a new visual generation method elevates GPT-style models beyond diffusion🚀 & Scaling laws observed📈
[](https://var.vision/demo)
[](https://arxiv.org/abs/2404.02905)
[](https://huggingface.co/FoundationVision/var)
[](https://paperswithcode.com/sota/image-generation-on-imagenet-256x256?tag_filter=485&p=visual-autoregressive-modeling-scalable-image)
Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction
NeurIPS 2024 Oral
## News
* **2024-09:** VAR is accepted as **NeurIPS 2024 Oral** Presentation.
* **2024-04:** [Visual AutoRegressive modeling](https://github.com/FoundationVision/VAR) is released.
## 🕹️ Try and Play with VAR!
We provide a [demo website](https://var.vision/demo) for you to play with VAR models and generate images interactively. Enjoy the fun of visual autoregressive modeling!
We also provide [demo_sample.ipynb](demo_sample.ipynb) for you to see more technical details about VAR.
[//]: # (
)
[//]: # (
)
## What's New?
### 🔥 Introducing VAR: a new paradigm in autoregressive visual generation✨:
Visual Autoregressive Modeling (VAR) redefines the autoregressive learning on images as coarse-to-fine "next-scale prediction" or "next-resolution prediction", diverging from the standard raster-scan "next-token prediction".
### 🔥 For the first time, GPT-style autoregressive models surpass diffusion models🚀:
### 🔥 Discovering power-law Scaling Laws in VAR transformers📈:
### 🔥 Zero-shot generalizability🛠️:
#### For a deep dive into our analyses, discussions, and evaluations, check out our [paper](https://arxiv.org/abs/2404.02905).
## VAR zoo
We provide VAR models for you to play with, which are on 
or can be downloaded from the following links:
| model | reso. | FID | rel. cost | #params | HF weights🤗 |
|:----------:|:-----:|:--------:|:---------:|:-------:|:------------------------------------------------------------------------------------|
| VAR-d16 | 256 | 3.55 | 0.4 | 310M | [var_d16.pth](https://huggingface.co/FoundationVision/var/resolve/main/var_d16.pth) |
| VAR-d20 | 256 | 2.95 | 0.5 | 600M | [var_d20.pth](https://huggingface.co/FoundationVision/var/resolve/main/var_d20.pth) |
| VAR-d24 | 256 | 2.33 | 0.6 | 1.0B | [var_d24.pth](https://huggingface.co/FoundationVision/var/resolve/main/var_d24.pth) |
| VAR-d30 | 256 | 1.97 | 1 | 2.0B | [var_d30.pth](https://huggingface.co/FoundationVision/var/resolve/main/var_d30.pth) |
| VAR-d30-re | 256 | **1.80** | 1 | 2.0B | [var_d30.pth](https://huggingface.co/FoundationVision/var/resolve/main/var_d30.pth) |
You can load these models to generate images via the codes in [demo_sample.ipynb](demo_sample.ipynb). Note: you need to download [vae_ch160v4096z32.pth](https://huggingface.co/FoundationVision/var/resolve/main/vae_ch160v4096z32.pth) first.
## Installation
1. Install `torch>=2.0.0`.
2. Install other pip packages via `pip3 install -r requirements.txt`.
3. Prepare the [ImageNet](http://image-net.org/) dataset
assume the ImageNet is in `/path/to/imagenet`. It should be like this:
```
/path/to/imagenet/:
train/:
n01440764:
many_images.JPEG ...
n01443537:
many_images.JPEG ...
val/:
n01440764:
ILSVRC2012_val_00000293.JPEG ...
n01443537:
ILSVRC2012_val_00000236.JPEG ...
```
**NOTE: The arg `--data_path=/path/to/imagenet` should be passed to the training script.**
5. (Optional) install and compile `flash-attn` and `xformers` for faster attention computation. Our code will automatically use them if installed. See [models/basic_var.py#L15-L30](models/basic_var.py#L15-L30).
## Training Scripts
To train VAR-{d16, d20, d24, d30, d36-s} on ImageNet 256x256 or 512x512, you can run the following command:
```shell
# d16, 256x256
torchrun --nproc_per_node=8 --nnodes=... --node_rank=... --master_addr=... --master_port=... train.py \
--depth=16 --bs=768 --ep=200 --fp16=1 --alng=1e-3 --wpe=0.1
# d20, 256x256
torchrun --nproc_per_node=8 --nnodes=... --node_rank=... --master_addr=... --master_port=... train.py \
--depth=20 --bs=768 --ep=250 --fp16=1 --alng=1e-3 --wpe=0.1
# d24, 256x256
torchrun --nproc_per_node=8 --nnodes=... --node_rank=... --master_addr=... --master_port=... train.py \
--depth=24 --bs=768 --ep=350 --tblr=8e-5 --fp16=1 --alng=1e-4 --wpe=0.01
# d30, 256x256
torchrun --nproc_per_node=8 --nnodes=... --node_rank=... --master_addr=... --master_port=... train.py \
--depth=30 --bs=1024 --ep=350 --tblr=8e-5 --fp16=1 --alng=1e-5 --wpe=0.01 --twde=0.08
# d36-s, 512x512 (-s means saln=1, shared AdaLN)
torchrun --nproc_per_node=8 --nnodes=... --node_rank=... --master_addr=... --master_port=... train.py \
--depth=36 --saln=1 --pn=512 --bs=768 --ep=350 --tblr=8e-5 --fp16=1 --alng=5e-6 --wpe=0.01 --twde=0.08
```
A folder named `local_output` will be created to save the checkpoints and logs.
You can monitor the training process by checking the logs in `local_output/log.txt` and `local_output/stdout.txt`, or using `tensorboard --logdir=local_output/`.
If your experiment is interrupted, just rerun the command, and the training will **automatically resume** from the last checkpoint in `local_output/ckpt*.pth` (see [utils/misc.py#L344-L357](utils/misc.py#L344-L357)).
## Sampling & Zero-shot Inference
For FID evaluation, use `var.autoregressive_infer_cfg(..., cfg=1.5, top_p=0.96, top_k=900, more_smooth=False)` to sample 50,000 images (50 per class) and save them as PNG (not JPEG) files in a folder. Pack them into a `.npz` file via `create_npz_from_sample_folder(sample_folder)` in [utils/misc.py#L344](utils/misc.py#L360).
Then use the [OpenAI's FID evaluation toolkit](https://github.com/openai/guided-diffusion/tree/main/evaluations) and reference ground truth npz file of [256x256](https://openaipublic.blob.core.windows.net/diffusion/jul-2021/ref_batches/imagenet/256/VIRTUAL_imagenet256_labeled.npz) or [512x512](https://openaipublic.blob.core.windows.net/diffusion/jul-2021/ref_batches/imagenet/512/VIRTUAL_imagenet512.npz) to evaluate FID, IS, precision, and recall.
Note a relatively small `cfg=1.5` is used for trade-off between image quality and diversity. You can adjust it to `cfg=5.0`, or sample with `autoregressive_infer_cfg(..., more_smooth=True)` for **better visual quality**.
We'll provide the sampling script later.
## License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## Citation
If our work assists your research, feel free to give us a star ⭐ or cite us using:
```
@Article{VAR,
title={Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction},
author={Keyu Tian and Yi Jiang and Zehuan Yuan and Bingyue Peng and Liwei Wang},
year={2024},
eprint={2404.02905},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```