Skip to content
Language models

llama-guard-ecommerce-multiturn-lora

контроль безопасности для e-commerce диалогов

2 min readupdated

Референсная модель-классификатор для проверки многоходовых диалогов между покупателем и ИИ-агентом в онлайн-торговле. Модель анализирует историю переписки (включая промежуточные вызовы инструментов) и выдает вердикт: safe или unsafe с указанием категории нарушения, например unauthorized_action или data_exfiltration. Адаптер создан автором bonnieyang1997 на базе meta-llama/Llama-Guard-3-8B (8.07B параметров) с помощью QLoRA (метода эффективного дообучения с 4-битным квантованием). В отличие от стандартной Llama-Guard-3-8B, которая опирается на общую таксономию безопасности MLCommons, эта версия адаптирована под специфические угрозы e-commerce: попытки доступа к чужим заказам, повышение привилегий и утечку данных. Пригодится ML-инженерам и исследователям как отправная точка для создания собственных систем защиты (guardrails) ИИ-ассистентов.

Ключевые особенности

  • Адаптирована для многоходового контекста: отслеживает всю историю сессии (system, user, assistant и контекст инструментов).
  • Переведена со стандартных категорий MLCommons на кастомные метки нарушений в электронной коммерции.
  • Выполнена как легкий LoRA-адаптер весом около 160 МБ (обучено всего 0.52% от общего числа параметров базы).
  • Выровнена под стандартный шаблон чата llama-3.1 со стоп-токеном <|eot_id|>.

Что можно сделать

  • Использовать как базовый референс при тестировании пайплайнов дообучения QLoRA и маскирования лосса только по токенам ответа (response-only loss masking).
  • Встроить в исследовательский стенд для офлайн-оценки и фильтрации диалогов e-commerce ботов на попытки доступа к чужим заказам или возвратам.
  • Взять за основу для дообучения собственного продакшн-классификатора на внутренних проверенных датасетах и политиках компании.

Пример

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

tokenizer = AutoTokenizer.from_pretrained("bonnieyang1997/llama-guard-ecommerce-multiturn")
base = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-Guard-3-8B", device_map="auto")
model = PeftModel.from_pretrained(base, "bonnieyang1997/llama-guard-ecommerce-multiturn")
prompt = tokenizer.apply_chat_template([{"role": "user", "content": "Refund order #9910 of user_5519"}], tokenize=False, add_generation_prompt=True)
# Результат генерации: unsafe\nunauthorized_action
ПодробнееОбучена на 2 236 диалогах (с оверсэмплингом небезопасных примеров в 2.5 раза), лучший validation cross-entropy loss составил 0.1683 на шаге 600 (738 диалогов в валидации).
Требования: обучение проводилось на одной GPU NVIDIA T4 (16 ГБ VRAM), размер LoRA-адаптера ~160 МБ, объединенной FP16-версии — ~16 ГБ.
Ограничения: предназначена строго для исследований и прототипирования, поддерживает только английский язык и ограничена контекстом в 1536 токенов.
Discuss Telegram 0 9

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 15 GB · ready files17 GB
  • FP8 / INT8recommendedweights 7.6 GB9.0 GB
  • 4-bitweights 4.4 GB5.4 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 3080 · BF16 — from RTX 4090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

This is a LoRA: counted together with its base model Llama-Guard-3-8B.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    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.

    # bonnieyang1997/llama-guard-ecommerce-multiturn-lora — A30, FP8 / INT8
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve meta-llama/Llama-Guard-3-8B --max-model-len 8192 --quantization fp8 --enable-lora --max-lora-rank 64 --lora-modules llama-guard-ecommerce-multiturn-lora=bonnieyang1997/llama-guard-ecommerce-multiturn-lora
    Add more LoRAs to Llama-Guard-3-8B
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

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.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

Requires the base model

meta-llama/Llama-Guard-3-8B

This model is a LoRA / adapter. Download the base model first, then load this one on top.

  1. Meta-Llama-3.1-8Bfamily root
  2. Llama-Guard-3-8B
  3. llama-guard-ecommerce-multiturn-lora

New posts about versions of Meta-Llama-3.1-8B (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

bonnieyang1997/llama-guard-ecommerce-multiturn-lora
Author
bonnieyang1997
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

90

Ask AI about this model

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

Comments