FSF-DMD

Pretrained weights and evaluation assets for Distribution Matching Distillation without Fake Score Network.

Code and instructions are available in the GitHub repository.

Contents

  • outputs/: Pretrained checkpoints.
  • buffers/vaes/: VAE checkpoints and latent statistics.
  • buffers/refs/: Reference files for FID evaluation.
Checkpoint Network Steps FID50K
2026.10.22KST20.04.12-fsfdmd-distill FSF-DMD-DiT-B/1 (VA-VAE) 3K 3.8942
2026.10.24KST16.38.59-fsfdmd-imf-official iMF-DiT-B/2 (SD-VAE) 1K 3.2964

Usage

Place the downloaded outputs and buffers directories at the top level of the code repository, then run the provided training or evaluation scripts.

TARGET=2026.10.22KST20.04.12-fsfdmd-distill bash scripts/eval.sh

Citation

@misc{kim2026distributionmatchingdistillationfake,
      title={Distribution Matching Distillation without Fake Score Network}, 
      author={Youngjoong Kim and Deokyeong Lee and Jaesik Park},
      year={2026},
      eprint={2605.19256},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2605.19256}, 
}
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Paper for YoungJoong/FSF-DMD