We currently have two ways of handling pre-trained models when building the ODELIA Docker image:
- cache in directory created inside local repository copy
- download during build to location in /tmp
We should not clutter the repository folder with such a cache and we should not rely on /tmp having enough storage space for a temporary download. Ideally, we should not rely on external resources being available when building images. Instead, we use consistent caching with a configurable location and the option to make caching optional.
We currently have two ways of handling pre-trained models when building the ODELIA Docker image:
We should not clutter the repository folder with such a cache and we should not rely on /tmp having enough storage space for a temporary download. Ideally, we should not rely on external resources being available when building images. Instead, we use consistent caching with a configurable location and the option to make caching optional.