Instructions to use niwz/Mini-Chinese-Phi3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use niwz/Mini-Chinese-Phi3 with Transformers:

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

pipe = pipeline("text-generation", model="niwz/Mini-Chinese-Phi3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, MiniPhi3

tokenizer = AutoTokenizer.from_pretrained("niwz/Mini-Chinese-Phi3")
model = MiniPhi3.from_pretrained("niwz/Mini-Chinese-Phi3", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use niwz/Mini-Chinese-Phi3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "niwz/Mini-Chinese-Phi3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "niwz/Mini-Chinese-Phi3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/niwz/Mini-Chinese-Phi3

SGLang

How to use niwz/Mini-Chinese-Phi3 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 "niwz/Mini-Chinese-Phi3" \
    --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": "niwz/Mini-Chinese-Phi3",
		"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 "niwz/Mini-Chinese-Phi3" \
        --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": "niwz/Mini-Chinese-Phi3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use niwz/Mini-Chinese-Phi3 with Docker Model Runner:
```
docker model run hf.co/niwz/Mini-Chinese-Phi3
```

Mini-Chinese-Phi3 / README.md

niuwz

modified: README.md

71d5f34 over 2 years ago

preview code

Raw History Blame Contribute Delete

1.13 kB

metadata

license: mit
datasets:
  - Skywork/SkyPile-150B
  - llm-wizard/alpaca-gpt4-data-zh
  - BelleGroup/train_2M_CN
  - BelleGroup/train_1M_CN
language:
  - zh
pipeline_tag: text-generation

Mini Chinese Phi3

Mini-Chinese-Phi3是一个基于phi3模型结构的小型对话模型，总参数量约0.13B，使用常见的中文语料进行预训练和微调。主要内容包括了

数据集的整理与简单清洗
中文词表预训练
基于phi3结构的模型预训练
基于预训练模型的指令微调（SFT），包括了全量微调和LoRA微调
基于指令微调模型的直接偏好优化（DPO）
模型评测 （待做）

项目中的所有训练过程均在两张3090显卡上进行，使用DeepSpeed框架和Flash Attention 2进行加速，预训练用时约40小时，SFT和DPO微调共用时约8小时。本项目是我在学习LLM过程中的一个简单实践，同时也希望能够帮助到同样初学大模型的小伙伴。

项目训练细节等已在Github上开源，欢迎大家提出宝贵意见和建议。项目地址