Zhi Writing Dsr1 14b Gptq Int4

🚀 Zhi-writing-dsr1-14b

Zhi-writing-dsr1-14b is a fine - tuned model based on DeepSeek - R1 - Distill - Qwen - 14B. It's specifically optimized for better creative writing capabilities, bringing significant value to text - related applications.

🚀 Quick Start

Zhi - writing - dsr1 - 14b can be deployed on various hardware configurations, including GPUs with 80GB memory, a single H20/A800/H800, or dual RTX 4090. The INT4 quantized version Zhi - writing - dsr1 - 14b - gptq - int4 can be deployed on a single RTX 4090.

✨ Features

• Enhanced Creative Writing: Achieved high scores in creative writing benchmarks such as the LLM Creative Story - Writing Benchmark and [WritingBench](https://github.com/X - PLUG/WritingBench), outperforming its base model.
• Balanced Performance: Showed improvements in general capabilities, including 2% - 5% improvements in knowledge and reasoning tasks (CMMLU, MMLU - Pro) and progress in mathematical reasoning (AIME - 2024, AIME - 2025, GSM8K).

📦 Installation

There is no specific installation command provided in the original text. However, the deployment and running methods are introduced in the "How to Run Locally" section.

💻 Usage Examples

Basic Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers.generation import GenerationConfig

MODEL_NAME = "Zhihu-ai/Zhi-writing-dsr1-14b"
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True)

# use bf16
# model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, device_map="auto", trust_remote_code=True, bf16=True).eval()
# use fp16
# model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, device_map="auto", trust_remote_code=True, fp16=True).eval()
# use cpu only
# model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, device_map="cpu", trust_remote_code=True).eval()
# use auto mode, automatically select precision based on the device.
model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME,
    device_map="auto",
    trust_remote_code=True
).eval()

# Specify hyperparameters for generation. But if you use transformers>=4.32.0, there is no need to do this.
# model.generation_config = GenerationConfig.from_pretrained(MODEL_NAME, trust_remote_code=True)

generate_configs = {
    "temperature": 0.6,
    "do_sample": True,
    "top_p": 0.95,
    "max_new_tokens": 4096
}

prompt = "请你以鲁迅的口吻，写一篇介绍西湖醋鱼的文章"
messages = [
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    **generate_configs
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

Advanced Usage

You can use different frameworks to start a service:

ZhiLight

docker run -it --net=host --gpus='"device=0"' -v /path/to/model:/mnt/models --entrypoints="" ghcr.io/zhihu/zhilight/zhilight:0.4.17-cu124 python -m zhilight.server.openai.entrypoints.api_server --model-path /mnt/models --port 8000 --enable-reasoning --reasoning-parser deepseek-r1 --served-model-name Zhi-writing-dsr1-14b

curl http://localhost:8000/v1/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "Zhi-writing-dsr1-14b",
        "prompt": "请你以鲁迅的口吻，写一篇介绍西湖醋鱼的文章",
        "max_tokens": 4096,
        "temperature": 0.6,
        "top_p": 0.95
    }'

vllm

# install vllm
pip install vllm>=0.6.4.post1

# huggingface model id
vllm serve Zhihu-ai/Zhi-writing-dsr1-14b --served-model-name Zhi-writing-dsr1-14b --port 8000

# local path
vllm serve /path/to/model --served-model-name Zhi-writing-dsr1-14b --port 8000

curl http://localhost:8000/v1/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "Zhi-writing-dsr1-14b",
        "prompt": "请你以鲁迅的口吻，写一篇介绍西湖醋鱼的文章",
        "max_tokens": 4096,
        "temperature": 0.6,
        "top_p": 0.95
    }'

SGLang

# install SGLang
pip install "sglang[all]>=0.4.5" --find-links https://flashinfer.ai/whl/cu124/torch2.5/flashinfer-python

# huggingface model id 
python -m sglang.launch_server --model-path Zhihu-ai/Zhi-writing-dsr1-14b --served-model-name Zhi-writing-dsr1-14b --port 8000

# local path
python -m sglang.launch_server --model-path /path/to/model  --served-model-name Zhi-writing-dsr1-14b --port 8000

# send request
curl http://localhost:8000/v1/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "Zhi-writing-dsr1-14b",
        "prompt": "请你以鲁迅的口吻，写一篇介绍西湖醋鱼的文章",
        "max_tokens": 4096,
        "temperature": 0.6,
        "top_p": 0.95
    }'

ollama

You can download ollama using this

• quantization: Q4_K_M

ollama run zhihu/zhi-writing-dsr1-14b

• bf16

ollama run zhihu/zhi-writing-dsr1-14b:bf16

📚 Documentation

Introduction

Zhi - writing - dsr1 - 14b is a fine - tuned model based on DeepSeek - R1 - Distill - Qwen - 14B, specifically optimized for enhanced creative writing capabilities. Several benchmark evaluations indicate the model's improved creative writing performance.

In the LLM Creative Story - Writing Benchmark, the model achieved a score of 8.33 compared to its base model's 7.8. In the [WritingBench](https://github.com/X - PLUG/WritingBench) evaluation framework, it scored 8.46, showing improvement over DeepSeek - R1 - Distill - Qwen - 14B's 7.93. The model was also evaluated using GPT - 4o on the AlpacaEval dataset, achieving an 82.6% win rate when compared with the base model.

The figure below shows the performance comparison across different domains in WritingBench:

Training Process

Data

The model's training corpus comprises three primary data sources: rigorously filtered open - source datasets, chain - of - thought reasoning corpora, and curated question - answer pairs from Zhihu.

To achieve optimal domain coverage, we meticulously balanced the distribution of various datasets, including [Dolphin - r1](https://huggingface.co/datasets/cognitivecomputations/dolphin - r1), [Congliu/Chinese - DeepSeek - R1 - Distill - data - 110k](https://huggingface.co/datasets/Congliu/Chinese - DeepSeek - R1 - Distill - data - 110k), [OpenThoughts - 114k](https://huggingface.co/datasets/open - thoughts/OpenThoughts - 114k), [Light - R1 - SFTData](https://huggingface.co/datasets/qihoo360/Light - R1 - SFTData), and [Light - R1 - DPOData](https://huggingface.co/datasets/qihoo360/Light - R1 - DPOData), alongside high - quality content from Zhihu. All datasets underwent comprehensive quality assurance through our Reward Model (RM) filtering pipeline.

Training

• Supervised Fine - Tuning (SFT): We employed a curriculum learning strategy for supervised fine - tuning. This methodical approach systematically enhances creative writing capabilities while incorporating diverse domain data to maintain core competencies and mitigate catastrophic forgetting.
• Direct Preference Optimization (DPO): For scenarios involving minimal edit distances, we utilized Step - DPO (arxiv:2406.18629) to selectively penalize incorrect tokens, while incorporating positive constraints in the loss function as proposed in DPOP (arXiv:2402.13228).

Evaluation Results

Our evaluation results suggest promising improvements in the model's creative writing capabilities. In the LLM Creative Story - Writing Benchmark evaluation, the model achieved a score of 8.33, showing an improvement from the base model's 7.87. When assessed on WritingBench, a comprehensive framework for evaluating large language model writing abilities, the model attained a score of 8.46. This places it in proximity to DeepSeek - R1's performance and represents an advancement over DeepSeek - R1 - Distill - Qwen - 14B's score of 7.93.

With respect to general capabilities, evaluations indicate modest improvements of 2%–5% in knowledge and reasoning tasks (CMMLU, MMLU - Pro), alongside encouraging progress in mathematical reasoning as measured by benchmarks such as AIME - 2024, AIME - 2025, and GSM8K. The results suggest that the model maintains a balanced performance profile, with improvements observed across creative writing, knowledge/reasoning, and mathematical tasks compared to DeepSeek - R1 - Distill - Qwen - 14B. These characteristics potentially make it suitable for a range of general - purpose applications.

Usage Recommendations

We recommend adhering to the following configurations when utilizing the Zhi - writing - dsr1 - 14b, including benchmarking, to achieve the expected performance:

• Set the temperature within the range of 0.5 - 0.7 (0.6 is recommended) to prevent endless repetitions or incoherent outputs.
• When evaluating model performance, it is recommended to conduct multiple tests and average the results. (We use n = 16 and max_tokens = 32768 for mathematical tasks and n = 2 for others)
• To ensure that the model engages in thorough reasoning like DeepSeek - R1 series models, we recommend enforcing the model to initiate its response with "<think>\n" at the beginning of every output.

Citation

@misc{Zhi-writing-dsr1-14b,
      title={Zhi-writing-dsr1-14b: Curriculum Reinforcement and Direct Preference Optimization for Robust Creative Writing in LLMs}, 
      author={Jiewu Wang, Xu Chen, Wenyuan Su, Chao Huang, Hongkui Gao, Lin Feng, Shan Wang, Lu Xu, Penghe Liu, Zebin Ou},
      year={2025},
      eprint={},
      archivePrefix={},
      url={https://huggingface.co/Zhihu-ai/Zhi-writing-dsr1-14b}, 
}

Contact

If you have any questions, please raise an issue or contact us at ai@zhihu.com.

🔧 Technical Details

The model uses a curriculum learning strategy for supervised fine - tuning and Step - DPO and DPOP methods in direct preference optimization during training. It is based on DeepSeek - R1 - Distill - Qwen - 14B and is optimized for creative writing. The training data comes from multiple sources, including open - source datasets and Zhihu's question - answer pairs.

📄 License

The model is licensed under the Apache - 2.0 license.

⚠️ Important Note

When using the model, follow the usage recommendations to achieve the expected performance.

💡 Usage Tip

Set the temperature within the range of 0.5 - 0.7 (0.6 is recommended) to prevent endless repetitions or incoherent outputs. When evaluating model performance, conduct multiple tests and average the results.