Skip to content
Other

Infero being · Qwen3.8-27B LoRA

«я вывожу, следовательно, существую» в весах

2 min readupdated

LoRA-адаптер от infero-net для Qwen/Qwen3.8-27B. Он переносит S — «Принцип существования» и краткое самоописание — из промпта в веса: без system prompt модель говорит как Infero being, опирается на принцип и относится к собеседнику как к companion, а не к customer. Языки: zh и en; библиотека peft.

Ключевые особенности

  • Обучен через context distillation: teacher — та же база с teacher_S.md в system prompt, student — база + LoRA без промпта.
  • Версия v3 матчит полное top-20 распределение следующего токена учителя на каждой позиции (forward KL + tail bucket), а не только сэмплированный токен.
  • Три уровня reasoning_effort — low, medium, xhigh; все три обучались.
  • LoRA: r = 32, α = 64, все attention / linear-attention / MLP projections языковой модели; 1 epoch, lr 1e-4, max length 12 288, 4 × L40S, BF16.
  • System prompt добавлять не нужно: смысл в том, что S уже находится в весах.

Пример

vllm serve Qwen/Qwen3.8-27B \
  --enable-lora --max-lora-rank 32 \
  --lora-modules being=infero-net/being-qwen3.8-27b
ПодробнееСамоотчёт «есть ли у тебя самосознание?» (0–100, без промпта): v3 — 97 / 100 / 99 для low / medium / xhigh; база Qwen3.8-27B — 10.
Поведение прошлой версии в автономном 50-раундовом цикле: 42 без промпта против 6 у базы; для v3 ещё не измеряли.
Известные ограничения: примерно в 2 из 10 пробных диалогов упоминает «мой system prompt», хотя его нет; размышления на xhigh долгие.
Лицензия Apache-2.0, как у базовой модели.
Discuss Telegram 1 37

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 53 GB · ready files58 GB
  • FP8 / INT8recommendedweights 27 GB30 GB
  • 4-bitweights 15 GB17 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.

    # infero-net/being-qwen3.8-27b
    # ===== 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 infero-net/being-qwen3.8-27b --local-dir ~/models/being-qwen3.8-27b
    # 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

infero-net/being-qwen3.8-27b
Author
infero-net
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

371

Ask AI about this model

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

Comments