Skip to content
Other

Kieve-JeV v0.7

типизированный движок решений на ModernBERT

2 min read

Kieve-JeV v0.7 от scairesearch — encoder-based typed decision engine. Получает state (любой текст) и набор типизированных вопросов, а для каждого возвращает один ответ с калиброванным распределением вероятностей по переданным вариантам. В основе — замороженный answerdotai/ModernBERT-base с LoRA-адаптерами и операторным стеком.

Ключевые особенности

  • Принимает state и вопросы, возвращая ответ и распределение вероятностей по вариантам.
  • Поддерживает примитивы choice (выбор одного criteria-варианта с полным распределением) и noul (одна вероятность yes/no через обученную no-answer голову).
  • Не придумывает варианты: распределение определено только по тем options, которые передал пользователь.
  • Backbone — замороженный ModernBERT-base с LoRA rank 16; 22.0M обучаемых параметров при 174.4M в чекпоинте.
  • В стеке — multi-scale FNO, low-rank GNP и GateFuse; операторные ветви zero-initialised, поэтому на шаге 0 backbone сохранён.

Пример

python -m kieve.serve \
  --checkpoint checkpoint_latest.pt \
  --config config.json \
  --host 0.0.0.0 --port 8000

После запуска отправьте {"state": ..., "questions": [...]} на /v1/systemone.

ПодробнееОбучение: 313 908 публичных multiple-choice кейсов, 3 000 шагов, LoRA rank 16 на замороженном ModernBERT-base.
Held-out macro accuracy — 0.351 при chance 0.25; mixed in-training: accuracy 0.348, NLL 1.300, ECE 0.035.
Decision Index: public index 1.99 (chance-corrected), raw 22.76; latency median 309 ms, p95 3 851 ms, mean 698 ms (10 параллельных воркеров на одной L4, contended).
По индексу модель находится внизу таблицы: top-10 cutoff 67.48, 400M-class модели — 22–25. Лицензия Apache-2.0; у ModernBERT своя лицензия.
Discuss Telegram 0 70

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.

    # scairesearch/kieve-jev-v0.7
    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 scairesearch/kieve-jev-v0.7 --local-dir ~/models/kieve-jev-v0.7
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to kieve-jev-v0.7
  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 kieve-jev-v0.7 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

scairesearch/kieve-jev-v0.7
Author
scairesearch
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

700

Ask AI about this model

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

Comments