## Basic Images for Inference

UHub provides a series of basic container images related to inference. Users can build custom images based on these basic images for deploying inference services.

### Image List

| Mirror Name       | Tag Information                             | Source Repository       |
|-------------------|---------------------------------------------|-------------------------|
| pytorch/pytorch   | 2.4.0-cuda11.8-cudnn9-runtime<br>2.4.0-cuda11.8-cudnn9-runtime | pytorch/pytorch         |
| nvidia/pytorch    | 24.07-py3<br>24.06-py3                       | nvcr.io/nvidia/pytorch  |
| vllm/vllm-openai  | v0.5.3<br>v0.5.2                             | vllm/vllm-openai        |
| jupyter/tensorflow-notebook | x86_64-ubuntu-22.04                     | jupyter/tensorflow-notebook |
| nvidia/cuda       | 12.5.1-cudnn-runtime-ubuntu20.04<br>12.5.1-cudnn-runtime-ubuntu22.04 | nvcr.io/nvidia/cuda     | 

### How to Obtain

All images in the above list have been synchronized to the repository: SCloudai, and the image address rule is as follows:

```
uhub.service.SCloud.cn/SCloudai/<image name>:tag
```

For example, to download the v0.5.3 version of the vllm/vllm-openai image, use the following command:

```
docker pull uhub.service.SCloud.cn/SCloudai/vllm/vllm-openai:v0.5.3
```