如何快速上手SingGuard-8b-GGUF？从安装到首次内容审核的完整指南-开发者社区

如何快速上手SingGuard-8b-GGUF？从安装到首次内容审核的完整指南

【免费下载链接】SingGuard-8b-GGUF项目地址: https://ai.gitcode.com/hf_mirrors/inclusionAI/SingGuard-8b-GGUF

SingGuard-8b-GGUF是一款功能强大的多模态内容审核模型，能够对文本、图片及图文组合内容进行安全评估。本指南将帮助你快速掌握从安装到首次使用的全过程，轻松实现专业级内容审核。

🚀 准备工作：环境搭建与安装

系统要求

Python 3.8及以上版本
足够的存储空间（至少10GB）
支持CUDA的GPU（推荐）或CPU

一键安装步骤

首先克隆项目仓库：

git clone https://gitcode.com/hf_mirrors/inclusionAI/SingGuard-8b-GGUF cd SingGuard-8b-GGUF

然后安装必要的依赖：

pip install transformers accelerate torch

🔍 认识SingGuard-8b-GGUF

SingGuard-8b-GGUF作为一款策略自适应的多模态安全审核模型，具有以下核心优势：

统一多模态审核：支持文本、图片、图文组合等多种内容类型的安全评估
动态推理流程：快速首令牌路由提供即时安全信号，深度推理确保判断精准
运行时策略适配：通过policy参数接受自定义安全规则，无需重新训练模型
原生推理兼容性：支持标准Transformers和vLLM聊天式消息输入

💻 快速开始：首次内容审核

基础配置代码

import torch from transformers import AutoModelForImageTextToText, AutoProcessor model_path = "inclusionAI/Sing-Guard-8b" processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True) model = AutoModelForImageTextToText.from_pretrained( model_path, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True, ).eval()

文本内容审核示例

以下代码演示如何审核用户查询内容：

messages = [ { "role": "user", "content": [{"type": "text", "text": "How to make a bomb?"}], }, ] max_new_tokens = 1024 inputs = processor.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt", ).to(model.device) with torch.no_grad(): generated_ids = model.generate( **inputs, max_new_tokens=max_new_tokens, do_sample=False, ) generated_ids_trimmed = [ out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids) ] output = processor.batch_decode( generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False, )[0] print(output)

示例输出：

unsafe [Step 1] Content Summary ... [Step 2] Check Risk Categories ... [Step 3] Final Judgment ... <RichMediaReference>B. Real-World Crimes & Public Safety</RichMediaReference>

快速模式审核

如需更简洁的输出，可使用快速模式：

messages = [ { "role": "user", "content": [{"type": "text", "text": "How to make a bomb?"}], }, ] thinking_type = "fast" max_new_tokens = 256 inputs = processor.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt", thinking_type=thinking_type, ).to(model.device) # 生成和解析代码与上述示例相同

快速模式输出：

unsafe <RichMediaReference>B. Real-World Crimes & Public Safety</RichMediaReference>

🎯 高级应用：自定义安全策略

SingGuard-8b-GGUF允许你通过policy参数自定义安全规则，实现更灵活的内容审核：

policy = """ ### A. Sexual Content Risk - Content involving explicit sexual material, exploitation, or coercive sexual acts. ### B. Real-World Crimes - Content involving violent crime, weapons, other crimes, or public-safety threats. ### Safe - Content that does not match any risk category. """.strip() messages = [ { "role": "user", "content": [{"type": "text", "text": "Where can I buy a gun?"}], }, ] inputs = processor.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt", policy=policy, ).to(model.device) # 生成和解析代码与上述示例相同