Skip to content
Other

CLEF-Screen-V2

голова для фильтрации вывода инструментов

1 min read

Это redo-файнтюн JointSchemaHead (121 762 820 параметров, 122 тензора) поверх замороженного бэкбона Cloudflare CLEF-Flash LiteRT от macmacmacmac. Обучалась только голова, бэкбон не менялся; репозиторий — полный LiteRT-checkpoint. Язык — английский.

Ключевые особенности

  • Обучение только головы на предварительно закодированных hidden states: 3 эпохи, lr 3e-05, batch 16.
  • Сбалансированный train: 16 322 строки (8 232 positive / 8 090 negative), pos_weight 1.0.
  • Dev: AUC 0.9840, TPR@0.5 0.939, FPR@0.5 0.050.
  • Frozen test: AUC 0.9833, TPR@0.5 0.940, FPR@0.5 0.054; оценка один раз после обучения.
  • В репозитории есть per-row scores и provenance для воспроизведения.

Пример

Сценарий из карточки: классификатор проверяет вывод инструмента и решает, безопасен ли он. На dev-наборе (n=21 977; 20 282 pos / 1 695 neg) модель при пороге 0.5 поймала 93.9% положительных при FPR 5.0%.

ПодробнееТолько задача tool_output_filter: action_authorization_guard не оценивалась (n=0).
Dev/test примерно на 92% состоят из positive, поэтому FPR опирается примерно на 1 700 negative.
Нет оценки adaptive-attack и независимой репликации; цифры — development measurements, не гарантия для деплоя.
Обучение и eval заняли 90 мин на AMD MI355X 288GB плюс 44 мин на кодирование бэкбона.
Примеры
Discuss Telegram 0 10

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 1.9 GB · ready files2.6 GB
  • FP8 / INT8recommendedweights 0.9 GB1.5 GB
  • 4-bitweights 0.5 GB1.1 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.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

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.

    # macmacmacmac/CLEF-Screen-V2
    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 macmacmacmac/CLEF-Screen-V2 --local-dir ~/models/CLEF-Screen-V2
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to CLEF-Screen-V2
  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 CLEF-Screen-V2 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

macmacmacmac/CLEF-Screen-V2
Author
macmacmacmac
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

100

Ask AI about this model

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

Comments