Commit History

  1. Install fusion-embedding from PyPI

    Verified signature
    120d15b
  2. Add Engram playground link

    Verified signature
    5f3eeb5
  3. Add Engram ecosystem section

    Verified signature
    f6d56b2

  1. Card: separate model citation (Tonmoy) from paper citation (all authors)

    Verified signature
    c1bbeca
  2. Cite the arXiv paper (2607.18666)

    Verified signature
    d94e3f9

  1. inference.py: native video preprocessing (no qwen-vl-utils)

    Verified signature
    b551ea8
  2. embed_video: native preprocessing (frame tensors and torchcodec path decode); drops the qwen-vl-utils dependency; bitwise-equal to the previous implementation and the raw base on identical inputs

    Verified signature
    d632d6b
  3. inference.py: embed_video (base's own video path; purely additive)

    Verified signature
    e63b6c0
  4. Add embed_video: the frozen base's own video path (qwen-vl-utils preprocessing, official defaults), gate-guarded like images; bitwise-verified against the base and the reference loader

    Verified signature
    f1a36c5

  1. Card: positioning alt text matches the trained-parameters axis

    Verified signature
    b82ccac
  2. Positioning figure: trained-parameters axis (matches the technical report)

    Verified signature
    415eb5c
  3. Remote code: batched embed_text_batch / embed_audio_batch entry points

    Verified signature
    e6d91bc
  4. AutoModel + trust_remote_code support (remote-code loading; existing loaders unchanged)

    Verified signature
    27e1dd7
  5. model card: positioning chart in cross-modal section

    Verified signature
    7898f53
  6. assets: unified-model positioning chart

    Verified signature
    ce32ae4
  7. model card: model overview figure

    Verified signature
    72d9220
  8. assets: model overview figure (brand palette)

    Verified signature
    cf18f7c

  1. Note the safetensors/param-count meaning

    Verified signature
    fdb9184
  2. Connector weights as safetensors (16.4M params)

    Verified signature
    9592473
  3. Add badges under the banner

    Verified signature
    d38ccf1
  4. Display banner image on the model card

    Verified signature
    ef18a87
  5. Add tagline banner to model card

    Verified signature
    67e0464
  6. MAEB: exact FSDKaggle2019 overlap numbers (608/4481 verified by Freesound id)

    Verified signature
    87f90d8
  7. MAEB: FSD2019Kaggle disclosure (anonymized filenames, disjointness unverifiable)

    Verified signature
    a65af61

  1. Add banner image

    Verified signature
    22f33a5
  2. Add banner image

    Verified signature
    bcbb63c
  3. Add banner to model card

    Verified signature
    2e9648e
  4. Add banner image

    Verified signature
    6bd45fd
  5. Restructure card: architecture up top, unified-model comparison leads evaluation

    Verified signature
    5020012
  6. MAEB(beta) results summary

    Verified signature
    15cd6b4
  7. Add Gemini Embedding 2 baseline to cross-modal comparison

    Verified signature
    49f04ad
  8. Add LanguageBind baseline to cross-modal comparison

    Verified signature
    ef02ac0
  9. v0.3 model card with three-version comparison

    Verified signature
    f263692
  10. v0.3 config

    Verified signature
    8e5ae97
  11. v0.3: AudioCaps in-domain fine-tune stage (a2t R@10 0.741, R@1 0.332)

    Verified signature
    e92d9df

  1. Model card: architecture diagram; fix bank size to 484K (v0.2)

    Verified signature
    bf1eb1a
  2. Add architecture diagram

    Verified signature
    abe5432

  1. Document version tags

    Verified signature
    b82df83
  2. from_pretrained: revision pinning

    Verified signature
    a815bcd
  3. v0.2 retrieval examples in model card

    Verified signature
    37f999f
  4. v0.2 audio-to-image retrieval gallery

    Verified signature
    11023b6
  5. Bold best-per-column only in comparison tables

    Verified signature
    c35a99c
  6. Bold v0.2 numbers in comparison tables

    Verified signature
    c72602c
  7. v0.2 model card with v0.1 comparison

    Verified signature
    889e950
  8. v0.2 config

    Verified signature
    efa97da
  9. v0.2: 484K-pair corpus checkpoint (AudioCaps 0.717, Clotho-ZS 0.448)

    Verified signature
    4ba4d39
  10. Add VGGSound attribution to gallery

    Verified signature
    afc6633
  11. Embed retrieval gallery in model card

    Verified signature
    054c8a8
  12. Add audio-to-image retrieval gallery

    Verified signature
    e085ecc
  13. Add full cross-modal metrics and qualitative retrieval examples

    Verified signature
    5874cec