Skip to content
Other

DOPEST 1.1 (31B)

вероятности вместо свободной генерации

2 min read

DOPEST 1.1 — типизированная модель решений: получает state (текст или JSON) и вопрос с закрытым набором вариантов, а возвращает калиброванные вероятности для каждого варианта за один forward pass. Это LoRA-адаптер (rank 32, последние 16 из 60 decoder-слоёв, 65M параметров) на базе google/gemma-4-31B-it. Поддерживает choice, yes/no и score.

Ключевые особенности

  • Для choice выдаёт распределение по именованным опциям, для yes/no — P(yes), для score — распределение по уровням и ожидаемый уровень.
  • Двунаправленное внимание внутри промпта на global-attention слоях; sliding-window слои остаются causal.
  • В репозитории есть calibration.json: температуры по типам вопросов и порог 0.801 для yes/no.
  • Опциональный второй проход: базовая модель пишет rationale до двух предложений, после чего опции читаются снова.
  • Второй проход включается только для неуверенных choice/score и ограничен бюджетом времени и стоимости.

Пример

uv run python -m dopest_model.serve --model google/gemma-4-31B-it --adapter dopest-studio/dopest-1.0-31b --calibration release/calibration.json --bidir full --reason --port 8000
curl -s localhost:8000/v1/systemone -H 'Content-Type: application/json' -d '{
  "state": "Order #4411 shipped on 3 March; arrived damaged on 9 March.",
  "questions": {"decision": {"type": "choice", "instructions": "Route this ticket.",
    "criteria": {"refund": "Refund or replacement", "shipping": "Delivery status"}}}}'
ПодробнееОфициальный результат v1.0 на JevBench: Capability 82.0 (Intelligence 73.9, Calibration 90.2).
На OpenDecisions один проход даёт 86.2, с --reason — 88.9 Capability.
Нужен GPU с 80 GB: bf16-веса около 62 GB плюс ~2 GB на кэш второго прохода.
Ограничения: только синтетические данные, слабее на длинной арифметике и неанглийских state; калибровку стоит проверять на своих данных.
Discuss Telegram 0 94

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.

    # dopest-studio/dopest-1.0-31b
    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 dopest-studio/dopest-1.0-31b --local-dir ~/models/dopest-1.0-31b
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to dopest-1.0-31b
  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 dopest-1.0-31b (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

dopest-studio/dopest-1.0-31b
Author
dopest-studio
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

940

Ask AI about this model

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

Comments