The (nvidia-only) quent-open db <run-id> feature currently has to infer which assets in a benchmark run is a quent asset. To avoid that, we could attach roles to a benchmark run:
"roles": [{"type": "quent", "filename": "asset-filename.zip"}]
Given a benchmark run with that role, quent-open could automatically use that asset without having to infer.
Assets also include media types. Perhaps we could standardize on an unofficial media type for quent and always use those.
The (nvidia-only)
quent-open db <run-id>feature currently has to infer which assets in a benchmark run is a quent asset. To avoid that, we could attachrolesto a benchmark run:Given a benchmark run with that role, quent-open could automatically use that asset without having to infer.
Assets also include media types. Perhaps we could standardize on an unofficial media type for quent and always use those.