Pre-built, GPU-optimized container images for running AI training and inference workloads on ACS clusters.
ACS AI container images are built for ACS clusters that use GPU‑accelerated instances and support all compatible GPU-accelerated instance types. Use these images to run training and inference for common AI workloads. More GPU-accelerated instance types will be added in future releases.
These images provide a fast‑start, open and compatible, performance‑optimized, and cost‑effective solution for AI training and inference at any scale.
Open and compatible
-
Built on the latest open‑source projects and technologies. Provides a community‑standard experience with no new learning curve.
Stable and reliable
-
Preinstalled training and inference frameworks, runtime libraries and tools, and the base environment are rigorously tested. Just add your models and data—ready to run. No environment setup required.
Optimized performance
-
Tuned for Alibaba Cloud infrastructure to deliver higher performance on the same hardware specifications.
Expert support
-
Each image includes up to one year of professional support.
-
Monthly rolling updates incorporate the latest improvements from the open-source community and Alibaba Cloud, ensuring you get new features and optimizations right away.