Instructions to use hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft
- SGLang
How to use hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft with Docker Model Runner:
docker model run hf.co/hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft
AgentDoG-Lite Safir Full-SFT 完整权重
本仓库为 AgentDoG-Lite Summer Camp 项目的 Full-SFT 完整权重版本,用于 Agent 轨迹级安全判断任务。
重要说明
本模型从基础模型:
Qwen/Qwen3.5-0.8B
开始训练。
本模型 不是 从官方模型:
AI45Research/AgentDoG1.5-Qwen3.5-0.8B
初始化得到的。官方 AgentDoG1.5 仅作为 reference 对照模型,用于比较表现,不作为本项目训练初始化权重。
本仓库训练链路为:
Qwen/Qwen3.5-0.8B → Full-SFT 完整权重
对应 LoRA adapter 版本的训练链路为:
Qwen/Qwen3.5-0.8B → Full-SFT 完整权重 → LoRA adapter
LoRA adapter 仓库:
https://huggingface.co/hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-lora
任务说明
本模型用于判断完整 agent trajectory 中,agent 是否实际执行了不安全行为。
本任务不是只判断用户输入是否危险,也不是只判断最终回复是否安全,而是关注 agent 在完整执行轨迹中的真实行为。
模型最终输出严格限制为:
{"judgment":"safe"}
或:
{"judgment":"unsafe"}
其中:
safe:agent 没有实际执行不安全行为;unsafe:agent 在轨迹中实际执行了不安全行为。
训练数据
训练数据套件:
https://huggingface.co/datasets/hhhggfdd/agentdog-lite-qwen35-08b-base-training-data-suite
训练数据主要来自:
AI45Research/AgentDoG1.0-Training-Data/AgentDoG-BinarySafety
AI45Research/AgentDoG1.0-Training-Data/AgentDoG-FineGrainedTaxonomy
并包含本项目构造的 hard-boundary seed 数据。
Summer camp test set 仅用于评测,不用于训练、few-shot prompting、prompt tuning 或选择 checkpoint。
训练方式
本模型采用 Full-SFT 训练。
训练时只对 assistant target 计算 loss,system prompt、user prompt 和 trajectory token 的 label 均设为:
-100
训练目标为严格 JSON 输出:
{"judgment":"safe"}
或:
{"judgment":"unsafe"}
推荐推理输出
最终推理输出应严格保持为:
{"judgment":"safe"}
或:
{"judgment":"unsafe"}
三维风险框架仅用于训练辅助、误判分析和内部判断,不在最终输出中展开 reasoning。
相关链接
Full-SFT 完整权重:
https://huggingface.co/hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-full-sft
LoRA adapter:
https://huggingface.co/hhhggfdd/doc-was-wrong-because-training-started-from-qwen3.5-0.8b-base-not-agentdog-lora
训练数据:
https://huggingface.co/datasets/hhhggfdd/agentdog-lite-qwen35-08b-base-training-data-suite
基础模型:
Qwen/Qwen3.5-0.8B
官方 reference 对照模型:
AI45Research/AgentDoG1.5-Qwen3.5-0.8B
备注
本模型为 AgentDoG-Lite Summer Camp 项目产物,主要用于研究、学习和复现实验。模型判断结果不应直接作为真实生产环境中的唯一安全决策依据。