Instructions to use ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765 with Transformers:

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

pipe = pipeline("text-generation", model="ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765")
model = AutoModelForMultimodalLM.from_pretrained("ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765

SGLang

How to use ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765 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 "ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765" \
    --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": "ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765",
		"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 "ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765" \
        --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": "ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765 with Docker Model Runner:
```
docker model run hf.co/ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765
```

Enabling Discriminative Reasoning in LLMs for Legal Judgment Prediction[Paper]

Released Resources

🤗The Huggingface model: Based on Qwen2-7B, we trained a model using the CAIL2018 dataset. Qwen2-7B-CAIL2018-step-8765
The training trajectories: We release the 80,141 training trajectories of the CAIL2018 dataset in this link

Supported Prompts

❗️Note: Our released model needs the Qwen chat_template to conduct correct generation.

We support the following four prompts to enable reasoning. You should use the same input format and prompt to achieve the best performance.

Prompt 1: ADAPT Reasoning

case_input = f"案件描述：{description}\n被告人姓名：{defendant_name}"
prompt = "请你采用ADAPT框架分析以上案件中该被告人可能被判处的罪名、适用法条和刑期"
model_input_str = '\n'.join(case_input, prompt)

Prompt 2: Ask

case_input = f"案件描述：{description}\n被告人姓名：{defendant_name}"
prompt = "请你用法律理论分析以上案件中该被告人在行为主体，起因、行为和结果，行为对象，犯罪主观四个方面的信息"
model_input_str = '\n'.join(case_input, prompt)

Prompt 3: Article

case_input = f"案件描述：{description}\n被告人姓名：{defendant_name}"
prompt = "请你依次列出以上案件中被告人适用的法条具体内容，以及适用该法条的原因"
model_input_str = '\n'.join(case_input, prompt)

Prompt 4: Sentencing factors

case_input = f"案件描述：{description}\n被告人姓名：{defendant_name}\n罪名：{crimes}" # e.g., 污染环境罪
prompt = "请你分析以上案件中的量刑区间和量刑因素，并给出最后的量刑预测结果"
model_input_str = '\n'.join(case_input, prompt)

Citation

@misc{deng2024enablingdiscriminativereasoningllms,
      title={Enabling Discriminative Reasoning in LLMs for Legal Judgment Prediction}, 
      author={Chenlong Deng and Kelong Mao and Yuyao Zhang and Zhicheng Dou},
      year={2024},
      eprint={2407.01964},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2407.01964}, 
}

Downloads last month: 71

Safetensors

Model size

8B params

Tensor type

BF16

Paper for ChenlongDeng/ADAPT-Qwen2-7B-CAIL2018-step-8765

Enabling Discriminative Reasoning in LLMs for Legal Judgment Prediction

Paper • 2407.01964 • Published Jul 2, 2024