bad-autoencoding / README.md
ivnle's picture
Upload README.md with huggingface_hub
942beb7 verified
|
raw
history blame
2.24 kB
metadata
license: apache-2.0
tags:
  - vision
  - ocr
  - compression
  - autoencoding

Bad Autoencoding - Model Checkpoints

Checkpoints for the paper: "Optical Context Compression Is Just (Bad) Autoencoding"

Ivan Lee, Cheng Yang, Taylor Berg-Kirkpatrick

Links

Available Checkpoints

Vision (base, 768x768)

Checkpoint Objective Hybrid Training CR PPL
vision_base_h0_recon Reconstruction 0 - 3.60 1.03
vision_base_h0_lm LM 0 Direct 3.60 5.08
vision_base_h0_lm_recon-init LM 0 From recon 3.60 5.06

Meanpool (w4s4)

Checkpoint Objective Hybrid Training CR PPL
meanpool_w4s4_h0_recon Reconstruction 0 - 3.97 1.04
meanpool_w4s4_h0_lm_recon-init LM 0 From recon 3.97 5.02

Naming Convention

{regime}_{config}_h{N}_{objective}[_recon-init]
Field Values Description
regime vision, conv1d, meanpool, text Compression architecture
config base/small/tiny/large, t500/t250, w4s4/w10s10, ctx525 Regime-specific config
h{N} h0, h100 Hybrid text tokens (0 = pure vision/compression)
objective recon, lm Training objective
recon-init (optional) LM initialized from reconstruction checkpoint

Model Details

  • Architecture: DeepSeek-OCR with trainable vision encoder
  • Encoder Status: Trained (not frozen)
  • Dataset: 510k samples from FineWiki

Usage

from huggingface_hub import hf_hub_download

# Download a specific checkpoint
checkpoint_path = hf_hub_download(
    repo_id="ivnle/bad-autoencoding",
    filename="vision_base_h0_lm/model.pt",
    repo_type="model"
)

Citation

@article{lee2024optical,
  title={Optical Context Compression Is Just (Bad) Autoencoding},
  author={Lee, Ivan and Yang, Cheng and Berg-Kirkpatrick, Taylor},
  journal={arXiv preprint arXiv:2512.03643},
  year={2024}
}