Instructions to use scisaga/qwen3-8b-instruct-lora-address-struct-cn with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use scisaga/qwen3-8b-instruct-lora-address-struct-cn with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="scisaga/qwen3-8b-instruct-lora-address-struct-cn")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("scisaga/qwen3-8b-instruct-lora-address-struct-cn")
model = AutoModelForCausalLM.from_pretrained("scisaga/qwen3-8b-instruct-lora-address-struct-cn")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use scisaga/qwen3-8b-instruct-lora-address-struct-cn with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "scisaga/qwen3-8b-instruct-lora-address-struct-cn"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "scisaga/qwen3-8b-instruct-lora-address-struct-cn",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/scisaga/qwen3-8b-instruct-lora-address-struct-cn

SGLang

How to use scisaga/qwen3-8b-instruct-lora-address-struct-cn 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 "scisaga/qwen3-8b-instruct-lora-address-struct-cn" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "scisaga/qwen3-8b-instruct-lora-address-struct-cn",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "scisaga/qwen3-8b-instruct-lora-address-struct-cn" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "scisaga/qwen3-8b-instruct-lora-address-struct-cn",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use scisaga/qwen3-8b-instruct-lora-address-struct-cn with Docker Model Runner:
```
docker model run hf.co/scisaga/qwen3-8b-instruct-lora-address-struct-cn
```

中文地址要素抽取 LoRA（Qwen3‑8B‑Instruct）

将中文地址文本直接生成 XML 标签串（如 <prov>…</prov><city>…</city>）。适配器采用 LoRA/QLoRA，与基座一起加载即可用于抽取省/市/区/道路/门牌/POI 等要素。

一、简介

任务：中文地址结构化（信息抽取 → 文本生成）。
形式：SFT 指令跟随；输入原始地址，输出 XML 标签串。
工程配套：提供 Web/REST 界面与 AMap 集成，以及 LoRA 数据与脚本。

二、代码与数据

代码仓库：Scisaga/addr-resolver（含 Web 端、API、LoRA 数据脚本、示例地址库等）。
- 关键目录/文件：lora/bio2sft.py、lora/build_sft_from_adm.py、func/amap_call.py、func/qwen_call.py、app.py、resolver.py、address.db（示例库）等。
系统能力：结合 高德地图 API 与 通义千问 进行解析/推理/定位；私有化地址库管理；提供 Docker 部署与前端页面。

三、训练

以 lora/ 目录产出的 sft.jsonl 为训练数据（BIO/BIES → SFT 一步式转换脚本已提供）。

CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node=2 lora/train_hf_qlora.py \
  --data lora/sft.jsonl --bf16

四、推理（docker）

docker run --gpus all --shm-size 1g -p 8080:80 \
  -v $PWD/qwen3-8b-instruct-lora-address-struct-cn:/data \
  ghcr.io/huggingface/text-generation-inference:latest \
  --model-id /data \
  --dtype bfloat16 \
  --num-shard 2 \
  --max-input-tokens 2048 --max-total-tokens 2304

五、许可

本仓库权重（适配器）与代码：CC BY 4.0（需署名、注明修改、不得施加额外法律/技术限制）。
基座 Qwen3‑8B‑Instruct：Apache‑2.0。如分发合并权重，需同时满足 Apache‑2.0（保留 LICENSE/NOTICE） 与 CC BY 4.0（署名/注明修改） 的要求。

推荐署名模板（在 README 或产品“关于”页中标注）：

Uses the model by Scisaga (CC BY 4.0). Based on Qwen/Qwen3‑8B‑Instruct (Apache‑2.0). Changes: LoRA fine‑tuning for Chinese address slot extraction.

六、致谢 / 变更

2025‑10‑26：首版精简模型卡，补充仓库链接与目录映射；许可调整为 CC BY 4.0。

Downloads last month: 3

Model tree for scisaga/qwen3-8b-instruct-lora-address-struct-cn

Base model

Qwen/Qwen3-8B-Base

Finetuned

Qwen/Qwen3-8B

Finetuned

(1634)

this model