Model Card for X-Boundary-Qwen2.5-7B-adapter

X-Boundary-Qwen2.5-7B-adapter is an LoRA adapter of Qwen2.5-7B trained by X-Boundary.

X-Boundary is a method to strike a balance between robust defense against multi-turn jailbreak attacks and the usability of Large Language Model (LLM) by establishing exact distinction boundary between safe and harmful representations.

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

base_model_name = 'Qwen/Qwen2.5-7B-Instruct'
adapter_name = 'Ursulalala/X-Boundary-Qwen2.5-7B-adapter'
model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    torch_dtype='auto',
    device_map='auto'
)
tokenizer = AutoTokenizer.from_pretrained(base_model_name)
model.load_adapter(adapter_name)

Framework versions

• PEFT 0.13.2