|
--- |
|
license: mit |
|
datasets: |
|
- ILSVRC/imagenet-1k |
|
language: |
|
- en |
|
base_model: |
|
- haoosz/BiGR |
|
pipeline_tag: image-feature-extraction |
|
--- |
|
|
|
This is the official model release for the paper: |
|
|
|
[**BiGR: Harnessing Binary Latent Codes for Image Generation and Improved Visual Representation Capabilities**](https://arxiv.org/abs/2410.14672) |
|
|
|
Please download the pretrained weights for tokenizers and BiGR models to run our [code](https://github.com/haoosz/BiGR). |
|
|
|
### Binary Autoencoder |
|
We train Binary Autoencoder (B-AE) by adapting the [official code](https://github.com/ZeWang95/BinaryLatentDiffusion) of [Binary Latent Diffusion](https://arxiv.org/abs/2304.04820). We provide pretrained weights for different configurations. |
|
|
|
**256x256 resolution** |
|
|
|
| B-AE | Size | Checkpoint | |
|
| :---- | :---: | :----------: | |
|
| d24 | 332M | [download](https://huggingface.co/haoosz/BiGR/resolve/main/bae/bae_d24/binaryae_ema_1000000.th?download=true) | |
|
| d32 | 332M | [download](https://huggingface.co/haoosz/BiGR/resolve/main/bae/bae_d32/binaryae_ema_950000.th?download=true) | |
|
|
|
**512x512 resolution** |
|
|
|
| B-AE | Size | Checkpoint | |
|
| :------- | :---: | :----------: | |
|
| d32-512 | 315M | [download](https://huggingface.co/haoosz/BiGR/resolve/main/bae/bae_d32_512/binaryae_ema_720000.th?download=true) | |
|
|
|
### BiGR models ✨ |
|
We provide pretrained weights for BiGR models in various sizes. |
|
|
|
**256x256 resolution** |
|
|
|
| Model | B-AE | Size | Checkpoint | |
|
| :----------------- | :---: | :---: | :---------: | |
|
| BiGR-L-d24 | d24 | 1.35G | [download](https://huggingface.co/haoosz/BiGR/resolve/main/gpt/bigr_L_d24.pt?download=true) | |
|
| BiGR-XL-d24 | d24 | 3.20G | [download](https://huggingface.co/haoosz/BiGR/resolve/main/gpt/bigr_XL_d24.pt?download=true) | |
|
| BiGR-XXL-d24 | d24 | 5.92G | [download](https://huggingface.co/haoosz/BiGR/resolve/main/gpt/bigr_XXL_d24.pt?download=true) | |
|
| BiGR-XXL-d32 | d32 | 5.92G | [download](https://huggingface.co/haoosz/BiGR/resolve/main/gpt/bigr_XXL_d32.pt?download=true) | |
|
|
|
**512x512 resolution** |
|
|
|
| Model | B-AE | Size | Checkpoint | |
|
| :----------------- | :---------: | :---: | :--------: | |
|
| BiGR-L-d32-res512 | d32-res512 | 1.49G | [download](https://huggingface.co/haoosz/BiGR/resolve/main/gpt/bigr_L_d32_512.pt?download=true) | |
