PAI Python SDK
English | 简体中文
PAI Python SDK是阿里云 机器学习平台 PAI(Platform for Artificial Intelligence) 提供的Python SDK，提供了更易用的HighLevel API，支持机器学习工程师简单地使用Python在PAI完成模型训练和部署，串联机器学习的流程。
🔧 安装
使用以下命令安装PAI Python SDK（支持Python版本 >= 3.8）：
python -m pip install pai
📖 文档
请通过访问 PAI Python SDK文档 或是查看 docs 目录下的文件获取SDK的详细文档，包括用户指南和API文档。
🛠 使用示例
- 提交自定义训练任务
以下代码演示了如何通过SDK提交一个自定义的训练作业:
from pai.estimator import Estimator
from pai.image import retrieve
est = Estimator(
# 获取PAI提供的最新PyTorch镜像
image_uri=retrieve(
framework_name="PyTorch", framework_version="latest"
).image_uri,
command="echo hello",
# 可选，指定source_dir上传你的训练代码：
# source_dir="./train_src",
instance_type="ecs.c6.large",
)
# 提交训练任务
est.fit()
print(est.model_data())
- 部署大语言模型
PAI提供了大量预训练模型，可以使用PAI Python SDK轻松部署：
from pai.model import RegisteredModel
# 获取PAI提供的QWen1.5-7b模型
qwen_model = RegisteredModel("qwen1.5-7b-chat", model_provider="pai")
# 部署模型
p = qwen_model.deploy(service_name="qwen_service")
# 调用服务
p.predict(
data={
"prompt": "What is the purpose of life?",
"system_prompt": "You are helpful assistant.",
"temperature": 0.8,
}
)
# PAI提供的大语言模型支持OpenAI API，可以通过openai SDK调用
openai_client = p.openai()
res = openai_client.chat.completions.create(
model="default",
max_tokens=1024,
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What is the purpose of life?"}
]
)
print(res.choices[0].message.content)
- 微调预训练模型
通过PAI提供的微调脚本，提交一个模型微调任务
from pai.model import ModelTrainingRecipe
training_recipe = ModelTrainingRecipe(
model_name="qwen2-0.5b-instruct",
model_provider="pai",
instance_type="ecs.gn6e-c12g1.3xlarge",
)
training_recipe.train(
inputs={
# 本地或是阿里云OSS上的数据路径(oss://<bucketname>/path/to/data)
"train": "<YourTrainingDataPath>"
}
)
通过访问PAI提供的示例仓库，可以了解更多使用示例：pai-examples
🤝 贡献代码
我们欢迎为PAI Python SDK贡献代码。请阅读 CONTRIBUTING 文件了解如何为本项目贡献代码。
📝 许可证
PAI Python SDK是由阿里云开发，并根据Apache许可证（版本2.0）授权使用。
📬 联系方式
如需支持或咨询，请在GitHub仓库中提交issue，或通过钉钉群联系我们：
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file alipai-0.4.13.tar.gz.
File metadata
- Download URL: alipai-0.4.13.tar.gz
- Upload date:
- Size: 379.2 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.1.0 CPython/3.8.18
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
c882f2b938143b798f53a91bd119f01f3f04746bf0c992e9e9b51a06bd72592a
|
|
| MD5 |
c2b92723a51eac63e2c66c2e6cd7e3df
|
|
| BLAKE2b-256 |
7b9d7daa1d31f648f0e049f22948d89a0abe4b98bd1100178de478d46af247e6
|
File details
Details for the file alipai-0.4.13-py3-none-any.whl.
File metadata
- Download URL: alipai-0.4.13-py3-none-any.whl
- Upload date:
- Size: 441.1 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/6.1.0 CPython/3.8.18
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
c8e4658e88e26f3d31a0ee3fc3c8cabebaf8ca8718edd0c5566460df5822ba13
|
|
| MD5 |
65e4e20c430680242200ab0c8ed6ee41
|
|
| BLAKE2b-256 |
8a5464ceba820fb0264170080ccd7b1be9be361b94913b9b6e4c3d0ab55d00f8
|
