Skip to content
Other

FactEmbed

факторизованные MLLM-эмбеддинги для composed image retrieval

1 min read

FactEmbed — исследовательская модель для Composed Image Retrieval (CIR): поиска целевого изображения по референсному изображению и тексту модификации. Построена на базе Qwen2.5-VL-7B-Instruct и разделяет понимание составного запроса и построение дискриминативных retrieval-эмбеддингов.

Ключевые особенности

  • Факторизованные латенты моделируют целевые визуальные модификации через intermediate target-imagination objective.
  • Используется factor-level contrastive loss и дистилляция декомпозированных представлений в более глубокий attention layer MLLM.
  • Для инференса сохраняется [EOS] token как глобальный эмбеддинг; mean-pooled факторизованные латенты реранжируют короткий список кандидатов в том же forward pass.
  • Чекпоинты обучены для FashionIQ и CIRR: stage3 — финальная модель для retrieval, stage1–2 — для воспроизводимости.

Пример

bash scripts/eval/eval_fashioniq.sh <path to fashioniq/stage3>
bash scripts/eval/eval_cirr.sh --model_path <path to cirr/stage3>
ПодробнееБенчмарки: FashionIQ и CIRR; в карточке заявлено превосходство над prior works, конкретные цифры не приведены.
Требования к VRAM/железу и ограничения в карточке не указаны.
Обучение: stage1 — target-imagination warm-up, stage2 — joint training (retrieval + imagination + distillation), stage3 — post-training с augmented triplets.
Discuss Telegram 1

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # WJ0830/factembed
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download WJ0830/factembed --local-dir ~/models/factembed
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to factembed
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of factembed (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

WJ0830/factembed
Author
WJ0830
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments