Skip to content
Other

SearchJev-27B

типизированный движок решений на Qwen3.8-27B

2 min readupdated

SearchJev-27B — LoRA-адаптер (rank 128, все linear projections) для Qwen/Qwen3.8-27B, который превращает базовую LLM в «System One» движок принятия решений. Он принимает state и типизированные вопросы (choice, noul да/нет, ordinal score) и за один forward pass возвращает вероятность каждого предложенного варианта. Текст модель не генерирует; адаптер merged into weights при загрузке, рядом есть сервер /v1/systemone и файлы калибровки.

Ключевые особенности

  • Один forward pass на вопрос: вероятности по всем вариантам, без генерации текста.
  • Типы вопросов: choice, noul (да/нет), ordinal score; температуры по типам — в calibration.json.
  • Адаптер rank 128 — усреднённый merge двух независимых LoRA: ΔW = 0.5·ΔW(stage 3) + 0.5·ΔW(round 2).
  • Готовый HTTP-сервер /v1/systemone и интеграция с decision-index kit.
  • Public suite: 140 178 из 140 178 запросов отвечены, без отказов и обрезаний.

Пример

pip install -r requirements.txt
hf download SearchJev/SearchJev-27B --local-dir SearchJev-27B && cd SearchJev-27B
python -m searchjev.server --model . --port 8000 --served-name searchjev-27b
python -m decision_index pipeline --engine http --option base_url=http://127.0.0.1:8000 \
    --option model=searchjev-27b --out runs/searchjev-27b
ПодробнееPublic index 73.52 (raw 79.71, breadth 72.97), completeness: 140 178/140 178.
Latency на NVIDIA H200: медиана 94 ms, 95-й перцентиль 864 ms.
Контекст — 262 144 токена; запрос сверх лимита получает HTTP 422. Сервер однопоточный, один запрос за раз.
Лицензия адаптера — CC BY-NC 4.0; базовая модель — Apache-2.0.
Discuss Telegram 0

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 55 GB · ready files61 GB
  • FP8 / INT8recommendedweights 29 GB33 GB
  • 4-bitweights 18 GB20 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

This is a LoRA: counted together with its base model Qwen3.8-27B.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # SearchJev/SearchJev-27B
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    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 Qwen/Qwen3.8-27B --local-dir ~/models/Qwen3.8-27B
    hf download SearchJev/SearchJev-27B --local-dir ~/models/SearchJev-27B
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Qwen3.8-27B
  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

Requires the base model

Qwen/Qwen3.8-27B

This model is a LoRA / adapter. Download the base model first, then load this one on top.

New posts about versions of Qwen3.8-27B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

SearchJev/SearchJev-27B
Author
SearchJev
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments