Instructions to use hiyouga/Llama-2-Chinese-13b-chat with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hiyouga/Llama-2-Chinese-13b-chat with Transformers:

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

pipe = pipeline("text-generation", model="hiyouga/Llama-2-Chinese-13b-chat")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("hiyouga/Llama-2-Chinese-13b-chat")
model = AutoModelForCausalLM.from_pretrained("hiyouga/Llama-2-Chinese-13b-chat")

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use hiyouga/Llama-2-Chinese-13b-chat with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hiyouga/Llama-2-Chinese-13b-chat"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hiyouga/Llama-2-Chinese-13b-chat",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/hiyouga/Llama-2-Chinese-13b-chat

SGLang

How to use hiyouga/Llama-2-Chinese-13b-chat 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 "hiyouga/Llama-2-Chinese-13b-chat" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hiyouga/Llama-2-Chinese-13b-chat",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "hiyouga/Llama-2-Chinese-13b-chat" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hiyouga/Llama-2-Chinese-13b-chat",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use hiyouga/Llama-2-Chinese-13b-chat with Docker Model Runner:
```
docker model run hf.co/hiyouga/Llama-2-Chinese-13b-chat
```

A bilingual instruction-tuned LoRA model of https://huggingface.co/meta-llama/Llama-2-13b-hf

Instruction-following datasets used: alpaca, alpaca-zh, open assistant
Training framework: LLaMA-Factory

Usage:

from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

tokenizer = AutoTokenizer.from_pretrained("hiyouga/Llama-2-Chinese-13b-chat")
model = AutoModelForCausalLM.from_pretrained("hiyouga/Llama-2-Chinese-13b-chat").cuda()
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

query = "晚上睡不着怎么办"
template = (
    "A chat between a curious user and an artificial intelligence assistant. "
    "The assistant gives helpful, detailed, and polite answers to the user's questions.\n"
    "Human: {}\nAssistant: "
)

inputs = tokenizer([template.format(query)], return_tensors="pt")
inputs = inputs.to("cuda")
generate_ids = model.generate(**inputs, max_new_tokens=256, streamer=streamer)

You could also alternatively launch a CLI demo by using the script in LLaMA-Factory

python src/cli_demo.py --template default --model_name_or_path hiyouga/Llama-2-Chinese-13b-chat

The model is trained using the web UI of LLaMA-Factory.

How to use this web UI: https://zhuanlan.zhihu.com/p/645010851

Loss curve:

Examples:

Q: 你是谁

A: 我是 <NAME>，一款由 <AUTHOR> 开发的人工智能助手。我可以回答各种问题，提供帮助和支持。

Q: 如何提高公司的销售额？

A: 以下是一些有助于提高公司销售额的想法：

改善销售流程：通过优化销售流程，销售人员可以更有效地与客户进行沟通，提高销售效率。
引入新产品：引入新产品可以吸引新客户，并增加现有客户的购买次数。
提高客户满意度：通过提高客户满意度，可以提高客户忠诚度和销售额。
营销活动：举办营销活动可以吸引新客户，并提高销售额。

Q: 给定下面的两个列表，使用Python解释器找到它们之间的相同元素。 List1：[1, 2, 3, 4, 5, 6] List2：[4, 5, 6, 7, 8, 9]

A: 以下是使用Python解释器找到两个列表中共同元素的示例：

# 列表1
lst1 = [1, 2, 3, 4, 5, 6]
# 列表2
lst2 = [4, 5, 6, 7, 8, 9]
# 找到共同元素
common_elements = list(set(lst1) & set(lst2))
print(common_elements)

此示例创建两个列表，然后使用Python中的集合操作符（&）找到它们共同的元素。最后，输出结果。在此示例中，共同元素为[4, 5, 6]。

Q: 将以下文本翻译成英语：我想去旅行并尽可能多地了解世界上的文化。

A: I want to travel and learn as much as possible about the cultures of the world.

Q: 解释以下缩写： JSON、CEO、PDF、HTML、NASA

A: 以下是缩写的解释：

JSON（JavaScript Object Notation）：一种用于序列化和交换数据的格式，常用于网络应用程序之间的通信。
CEO（Chief Executive Officer）：公司的最高负责人，负责公司的总体运营和管理。
PDF（Portable Document Format）：一种格式，可以将文档保存为一个无需特殊软件即可打开和查看的电子文件。
HTML（Hypertext Markup Language）：一种用于创建网页的标记语言，它包含用于格式化文本和添加元素的标记。
NASA（National Aeronautics and Space Administration）：美国的国家航空航天局，负责美国的宇宙计划和太空探索活动。

Downloads last month: 21

hiyouga
/

Llama-2-Chinese-13b-chat

Datasets used to train hiyouga/Llama-2-Chinese-13b-chat