#Which are the non-dev ROCm docker images?

1 messages · Page 1 of 1 (latest)

tight cape
#

The recommended base ROCm docker image is the dev image, based on rocm-dev. Is there a more lightweight image for runtime only?

#

Note I'm not using a Python framework, so I don't want e.g. the JAX, pytorch or tf images

pure snowBOT
#

While our rocm/dev images are the standard for building applications, we offer several alternatives for more specialized or lightweight needs:

  • rocm/rocm-terminal: This is a small image containing the prerequisites to build HIP applications, though it does not include any libraries. [rocm.docs...]
  • Framework-Specific Images: For production or runtime environments, we provide pre-built images for major frameworks that include the necessary ROCm backends:
  • rocm/pytorch: Pre-built PyTorch environment. [rocm.blogs...]
  • rocm/tensorflow: Pre-built TensorFlow environment. [rocm.blogs...]
  • rocm/jax: Pre-built JAX environment. [rocm.blogs...]
  • rocm/vllm: This is a ROCm-optimized build specifically for LLM inference and benchmarking, making it ideal for deployment scenarios. [rocm.blogs...]
#

Using these containers ensures your applications run consistently across different systems and ROCm versions while providing isolation for multiple applications on the same system. [rocm.blogs...] [rocm.blogs...]

https://rocm.blogs.amd.com/_images/images-2025-01-09-containers.webp "Getting started with AMD ROCm containers: from base images to custom solutions"

Getting started with AMD ROCm containers: from base images to custom solutions (Image from rocm.blogs.amd.com)

References:
• Running ROCm Docker containers — ROCm installation
• Getting started with AMD ROCm containers: from bas