Skip to content
Other

Amber 0.4B v2

компактный движок решений на 397,9 млн параметров

1 min read

Amber 0.4B v2 — типизированный decision engine от MLukauskas. На вход получает state и вопросы с вариантами (choice или да/нет noul), а на выходе даёт вероятность каждого варианта. Основа — cross-encoder/ettin-reranker-400m-v1 с дообученными весами и небольшим option head.

Ключевые особенности

  • 397,9M параметров, архитектура cross-encoder: один проход по инструкциям, состоянию и всем вариантам сразу.
  • Для каждого варианта считается softmax-вероятность; для noul ответ — P(true).
  • До 32 768 токенов на вопрос; длинные вопросы обрабатываются отдельно с sliding-window attention.
  • Средняя задержка 41,0 мс, медиана 19,2 мс на одной RTX 3080 Ti в bf16.
  • Есть варианты 17M, 32M, 68M, 150M и 0.4B.

Пример

pip install -r requirements.txt
PYTHONPATH=. python -m decision_index run --engine amber_uni_engine:UniEngine \
    --suite-dir <suite-0.3> --out runs/amber-0.4b-v2 --compact
ПодробнееPublic index — 47,42 (raw 60,00) в Decision Index 0.3: 140 178 запросов без пропусков и обрезаний.
Слабое место — Knowledge & Reasoning: маленький encoder без генерации близок к случайному уровню на задачах с мировыми знаниями и многошаговой арифметикой.
В основном английский; вероятности откалиброваны на тренировочном миксе, пороги лучше перепроверять на своих данных.
Веса — CC BY-NC 4.0 из-за некоммерческих источников в данных, базовая модель — Apache-2.0.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.7 GB · ready files1.3 GB
  • FP8 / INT8recommendedweights 0.4 GB0.9 GB
  • 4-bitweights 0.2 GB0.7 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.

    # MLukauskas/amber-0.4b-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 MLukauskas/amber-0.4b-v2 --local-dir ~/models/amber-0.4b-v2
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to amber-0.4b-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

  1. ettin-encoder-400mfamily root
  2. ettin-reranker-400m-v1
  3. amber-0.4b-v2

New posts about versions of ettin-encoder-400m (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 7All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 1All on Hugging Face

About the model

MLukauskas/amber-0.4b-v2
Author
MLukauskas
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