Install OSS Connector - Object Storage Service - Alibaba Cloud Documentation Center

Install the OSS Connector for AI/ML to efficiently access and store datasets in OSS for your PyTorch training jobs.

Prerequisites

Operating system: Linux x86-64
glibc: >=2.17
Python: 3.8–3.13
PyTorch: >=2.0
To use the OSS Checkpoint feature, your Linux kernel must support userfaultfd.

Note
For example, on an Ubuntu system, run the sudo grep CONFIG_USERFAULTFD /boot/config-$(uname -r) command to check if your Linux kernel supports userfaultfd. If the command returns CONFIG_USERFAULTFD=y, your kernel supports userfaultfd. If it returns CONFIG_USERFAULTFD=n, it does not, and you cannot use the OSS Checkpoint feature.

Procedure

The following example shows how to install the OSS Connector for AI/ML for Python 3.12.

On your Linux operating system or in a container built from a Linux image, run the pip3.12 install osstorchconnector command to install the OSS Connector for AI/ML.
```
pip3.12 install osstorchconnector
```

Run pip3.12 show osstorchconnector to verify the installation.

pip3.12 show osstorchconnector

A successful installation displays the version information for osstorchconnector. The output is similar to the following:

Name: osstorchconnector
Version: 1.0.0rc1
Summary: OSS connector for AI/ML
Home-page: 
Author: 
Author-email: 
License: 
Location: /usr/local/lib/python3.12/dist-packages
Requires: torch
Required-by:

Next steps

After installation, configure the connector's access credentials and settings. This allows the connector to communicate with OSS and initialize correctly. For instructions, see Configure OSS Connector for AI/ML.