Skip to content
Computer vision

PURSE splitter (Qwen3.5 9B)

LoRA-модель для разбивки бюджета

2 min read

PURSE splitter — исследовательский файнтюн Qwen/Qwen3.5-9B (9,41 млрд параметров) от guineapig. Он читает описание проекта с бюджетом, вариантами моделей/агентов и таблицей успешности, а затем возвращает JSON: какую опцию выбрать для каждой задачи или пропустить её. Это не общий ассистент, а LoRA-копия разбивки knapsack-солвера из бенчмарка PURSE (HASHIRU).

Ключевые особенности

  • Отвечает только JSON-объектом: для каждой из 8 задач — модель, внешний агент или skip.
  • Запуск с выключенным thinking и temperature 0, как при обучении.
  • Обучена на 3000 промптах PURSE; цель — копировать разбивку knapsack-солвера.
  • LoRA rank 16, alpha 32, dropout 0.05; обучение на одной H100 80GB около 33 минут.
  • Работает в одном рынке: шесть опций, четыре семейства задач, уровни 1–5.

Пример

from openai import OpenAI
client = OpenAI(base_url="http://127.0.0.1:8000/v1", api_key="none")
reply = client.chat.completions.create(
    model="guineapig/purse-splitter-qwen3.5-9b",
    messages=[{"role": "user", "content": prompt}],
    temperature=0, max_tokens=512,
    extra_body={"chat_template_kwargs": {"enable_thinking": False}},
)
split = reply.choices[0].message.content
ПодробнееОценка на 20 новых проектах: 0.663 / 0.750 / 0.820 / 0.879 при τ 0.25 / 0.5 / 1 / 2; knapsack-солвер — 0.715 / 0.788 / 0.842 / 0.902.
Merged-чекпоинт в bfloat16 — около 18.8 GB, адаптер — около 112 MB. Обучение заняло ~33 минуты на одной H100 80GB.
Ограничения: только шесть опций, четыре семейства задач и восемь задач на проект; при других ценах/опциях, вероятно, нужен новый трейнинг.
Discuss Telegram 0 37

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.

    # guineapig/purse-splitter-qwen3.5-9b
    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 guineapig/purse-splitter-qwen3.5-9b --local-dir ~/models/purse-splitter-qwen3.5-9b
    # The model is in ~/models. How to run it is in the model card (link above).
  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 purse-splitter-qwen3.5-9b (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

guineapig/purse-splitter-qwen3.5-9b
Author
guineapig
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

370

Ask AI about this model

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

Comments