問題文
An operations engineer must give three different project teams the same version of a deep learning framework, with the matching numerical libraries already validated together, on servers whose base operating systems differ. The engineer wants to avoid rebuilding the software on each host. Which approach matches the intent of NVIDIA's container catalog?
選択肢
- Pull a pre-built container image that already bundles the framework and its validated libraries, and run it on each host.
- Copy the working directory from one host to the others with a file transfer tool so the binaries are identical.
- Compile the framework from source on each host so that it links against the libraries that host already has.
- Install the framework with the operating system's package manager on every host, which resolves the library versions automatically.