Skip to content
Computer vision

Agastya Decision RAG 2B

2B-модель для решений в RAG

2 min readupdated

Это файнтюн Qwen3.5-2B (2.27B параметров) от amartya16 с головой Unsloth Clef для typed decision inference. Модель отвечает на типизированные вопросы о переданном состоянии и возвращает вероятности; её цель — маршрутизация и выбор внутри RAG-системы, управляемой оглавлением.

Ключевые особенности

  • Поддерживает вопросы choice, score и noul (boolean) — с вероятностями ответов.
  • Для ToC-guided RAG: выбрать следующий раздел или решить, нужен ли ещё один retrieval-шаг.
  • Для typed decision inference нужен FastDecisionModel из Unsloth; обычная загрузка через Transformers не открывает этот API.
  • Обучение: QLoRA на LocalLLaMA/typed-decisions, soft cross-entropy, 34 шага, 2 эпохи.
  • В репозитории есть handler.py для HF Inference Endpoints, но деплой и совместимость с GPU ещё не тестировались.

Пример

from unsloth import FastDecisionModel
model, tok = FastDecisionModel.from_pretrained("amartya16/Agastya-RAG-2B", max_seq_length=2048, load_in_4bit=True)
FastDecisionModel.for_inference(model)
answers = FastDecisionModel.predict(model=model, tokenizer=tok,
    state="Question: When is an appeal due? ToC: 1. Scope; 2. Filing deadlines; 3. Appeals.",
    questions={"next_section": {"type": "choice", "instructions": "Select the best section to read next.", "criteria": {"deadlines": "Section 2", "appeals": "Section 3"}}})
print(answers)
ПодробнееНа 150 held-out typed decisions: 66.0% accuracy и ECE 0.0948; у базовой модели — 37.33% и 0.0967.
Файл backbone ~4.55 ГБ, голова ~61 МБ; в decision-конфиге max_len: 4096.
ToC-маршрутизация отдельно не оценена; модель не индексирует документы, не ищет пассажи и не цитирует источники.
Лицензия Apache-2.0.
Discuss Telegram 1 31

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 4.2 GB · ready files5.2 GB
  • FP8 / INT8recommendedweights 2.1 GB2.8 GB
  • 4-bitweights 1.2 GB1.8 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.

    # amartya16/Agastya-RAG-2B
    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 amartya16/Agastya-RAG-2B --local-dir ~/models/Agastya-RAG-2B
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Agastya-RAG-2B
  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. Qwen3.5-2B-Basefamily root
  2. Qwen3.5-2B
  3. Qwen3.5-2B
  4. Agastya-RAG-2B

New posts about versions of Qwen3.5-2B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 10All on Hugging Face
Merges 2All on Hugging Face

About the model

amartya16/Agastya-RAG-2B
Author
amartya16
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

311

Ask AI about this model

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

Comments

Other

QEV-0.8B

QEV-0.8B