Text Generation
Transformers
Safetensors
Chinese
qwen3_5_text
qwen
dora
orpo
chinese
creative-writing
web-novel
conversational
Instructions to use wcn123/Qwen3.5-27B-WebNovel-Writer-zh with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use wcn123/Qwen3.5-27B-WebNovel-Writer-zh with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="wcn123/Qwen3.5-27B-WebNovel-Writer-zh") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("wcn123/Qwen3.5-27B-WebNovel-Writer-zh") model = AutoModelForCausalLM.from_pretrained("wcn123/Qwen3.5-27B-WebNovel-Writer-zh") 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 wcn123/Qwen3.5-27B-WebNovel-Writer-zh with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "wcn123/Qwen3.5-27B-WebNovel-Writer-zh" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "wcn123/Qwen3.5-27B-WebNovel-Writer-zh", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/wcn123/Qwen3.5-27B-WebNovel-Writer-zh
- SGLang
How to use wcn123/Qwen3.5-27B-WebNovel-Writer-zh 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 "wcn123/Qwen3.5-27B-WebNovel-Writer-zh" \ --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": "wcn123/Qwen3.5-27B-WebNovel-Writer-zh", "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 "wcn123/Qwen3.5-27B-WebNovel-Writer-zh" \ --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": "wcn123/Qwen3.5-27B-WebNovel-Writer-zh", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use wcn123/Qwen3.5-27B-WebNovel-Writer-zh with Docker Model Runner:
docker model run hf.co/wcn123/Qwen3.5-27B-WebNovel-Writer-zh
Qwen3.5-27B-WebNovel-Writer-zh
基于 Qwen3.5-27B-text 的中文网文写作风格微调模型,目标为去除 AI 味、向网文原著白描风格对齐。
基座模型 Qwen3.5-27B-text 为作者从 Qwen3.5-27B 抽离纯文本权重的自定义版本,非官方发布。另有 Q4_K_M 量化版本。
训练
两阶段微调:
- SFT:20,320 条中文网文正文,DoRA 16bit,学习叙事节奏与白描风格
- ORPO:8,000 条偏好对,强化去 AI 味倾向
| 参数 | 值 |
|---|---|
| 基座 | Qwen3.5-27B-text |
| 方法 | DoRA 16bit |
| LoRA rank / alpha | 64 / 128 |
| Target modules | 12(full attention + MLP) |
| SFT LR | 3e-5 |
| ORPO LR / beta | 5e-6 / 0.15 |
使用建议
System prompt:
你是一位中文西幻网文写作助手,擅长创作高质量的小说正文。请根据用户的指令完成写作任务。
训练数据包含以下任务格式,按对应格式输入可获得最佳效果:
正文增强(beat → 完整场景)
任务:正文增强
目标:让这段更像成熟作者写出的西幻正文
输入类型:场景beat
增强强度:中(从场景beat到完整场景)
长度目标:2.0x~3.5x
重点:节奏铺排、心理层次、环境渲染、微动作衔接
限制:严格保留骨架中的全部事实、人名、地名、术语、事件顺序与结果。不新增原文没有的信息。
素材:
[场景beat内容]
骨架扩写(骨架 → 完整场景)
请将下面文本增强为更自然的小说正文。
输入类型:事件骨架
增强强度:高(从纯骨架到完整场景)
长度目标:3.5x~6.0x
重点:从骨架扩写完整场景:叙事结构、节奏铺排、感官填充、对话还原
要求:严格保留骨架中的全部事实、人名、地名、术语、事件顺序与结果。不新增原文没有的信息。
文本:
[事件骨架内容]
场景扩写(描述 → 正文)
任务:场景扩写
场景描述:[场景描述]
要求:西幻风格,注重场景感和节奏,保持人物行为合理。
请写出完整的小说场景。
正文润色
任务:正文润色
要求:提升文笔、优化节奏、保留原意
原文:
[待润色文本]
推荐推理参数:temperature=0.7,top_p=0.90,repetition_penalty=1.10。
数据
中文网文正文语料,版权归原作者所有,数据集不公开。
协议
遵循 Qwen License,仅供研究与个人使用。
- Downloads last month
- 21