Open-weight model
Frechet-Distributional-Decoder-Alignment
by Zhu jiajunzhu/Frechet-Distributional-Decoder-Alignment
Frechet-Distributional-Decoder-Alignment is an open-weight model from Zhu. Its published files total 13.7 GB.
Decoder weights of FDDA. FDDA post-trains the tokenizer decoder of a frozen latent generative model with a Fréchet distance loss on generated samples.
Model Card
Decoder weights of FDDA. FDDA post-trains the tokenizer decoder of a frozen latent generative model with a Fréchet distance loss on generated samples. Code, installation and evaluation: Each checkpoint stores the tokenizer state dict with the adapted decoder ({"model", "epoch"}) and is used with the generator in --generatorname. ImageNet 256x256, 50,000 generated samples. gFD r6 is the mean over six representation spaces (Inception-v3, ConvNeXt-v2, DINOv2, MAE, SigLIP2, CLIP) of FD divided by the FD of the ImageNet validation set. beforegeneratorside/ /: decoders adapted to the
Excerpt from the card by Zhu.
Identity and Version
- Repository
- jiajunzhu/Frechet-Distributional-Decoder-Alignment
- Publisher
- Zhu
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 6e7eb82df5876621ef88d607253185d30be268bd
- First published
- 2026-10-04
- Last updated
- 2026-10-04
Files and Weights
30 files, 13.7 GB in total. The weights are 28 files totalling 13.7 GB in npz, pt.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| after_generator_side/gigatok/gigatok-B_256-fdpt.pt | Weights | 544.8 MB | 89716019abd7 |
| after_generator_side/imf/imf-B_256-fdsim.pt | Weights | 334.7 MB | a7c30d083137 |
| after_generator_side/imf/imf-L_256-fdsim.pt | Weights | 334.7 MB | 0beea1f89667 |
| after_generator_side/imf/imf-XL_256-fdsim.pt | Weights | 334.7 MB | 1642e41ed775 |
| after_generator_side/llamagen/llamagen-B_256-fdpt.pt | Weights | 287.7 MB | 65dbb5bd38be |
| after_generator_side/llamagen/llamagen-L_256-fdpt.pt | Weights | 287.7 MB | 4637241f40db |
| after_generator_side/titok/titok-B64-fdpt.pt | Weights | 819.4 MB | 6621d5a56b46 |
| after_generator_side/titok/titok-L32-fdpt.pt | Weights | 2.6 GB | 577a6f7ade72 |
| after_generator_side/var/var-d16_256-fdpt.pt | Weights | 436.1 MB | 259594ed2a24 |
| after_generator_side/var/var-d20_256-fdpt.pt | Weights | 436.1 MB | 4ca15748715d |
| after_generator_side/var/var-d24_256-fdpt.pt | Weights | 436.1 MB | 2193bcd75666 |
| before_generator_side/gigatok/gigatok-B_256.pt | Weights | 544.8 MB | 6dc9810663fc |
| before_generator_side/imf/imf-B_256.pt | Weights | 334.7 MB | 4693338f6b38 |
| before_generator_side/imf/imf-L_256.pt | Weights | 334.7 MB | c269c52197e2 |
| before_generator_side/imf/imf-XL_256.pt | Weights | 334.7 MB | f0805d732808 |
| before_generator_side/llamagen/llamagen-B_256.pt | Weights | 287.7 MB | 4bc8ba591cfc |
| before_generator_side/llamagen/llamagen-L_256.pt | Weights | 287.7 MB | adb6db30e771 |
| before_generator_side/titok/titok-B64.pt | Weights | 819.4 MB | 2e337d479712 |
| before_generator_side/titok/titok-L32.pt | Weights | 2.6 GB | de2fc5479056 |
| before_generator_side/var/var-d16_256.pt | Weights | 436.1 MB | 104163ed0c93 |
| before_generator_side/var/var-d20_256.pt | Weights | 436.1 MB | 6c4c6920ee21 |
| before_generator_side/var/var-d24_256.pt | Weights | 436.1 MB | a7eea1583380 |
| reference_stats/reconstruction/fid_stats/rec_convnext_in256_t224_stats.npz | Weights | 8.4 MB | f989bbc0a1a9 |
| reference_stats/reconstruction/fid_stats/rec_inception_stats.npz | Weights | 33.6 MB | 758b856810c0 |
| reference_stats/reconstruction/fid_stats/rec_vit_large_patch14_clip_224_openai_in256_t256_stats.npz | Weights | 16.8 MB | e825c2ee98cc |
| reference_stats/reconstruction/fid_stats/rec_vit_large_patch14_dinov2_lvd142m_in256_t256_stats.npz | Weights | 16.8 MB | dc5eceb142a0 |
| reference_stats/reconstruction/fid_stats/rec_vit_large_patch16_224_mae_in256_t224_stats.npz | Weights | 16.8 MB | c86c22c6531b |
| reference_stats/reconstruction/fid_stats/rec_vit_so400m_patch16_siglip_256_v2_webli_in256_t224_stats.npz | Weights | 21.3 MB | c65b821543c3 |
| README.md | Documentation | 6.2 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 13.7 GB
Released by Zhu through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 13.7 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.