You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We would like to be able to generalize models and organize them in sets:
Categories (e.g. "en-jp text translation", "instruct")
Versions (e.g. ">= 1.3.*")
Suggestion: support "tagging" models in the cache with a set of searchable string values, using a metadata index.
Caches could then be searched for models with certain tags. Searching should allow for comparisons/regexes so some tags can represent versions.
This is similar to how hashtags are used for github topics, tweets, etc. Each model could have zero or more tags.
May also want to consider extending this to support performance/cost metadata (e.g. number of parameters, quantization level, etc). This can also be used to support "friendly" names for models. Tags should be included in the hash to avoid using them for cross-site communication and protecting them from modification.
Assuming tagging was implemented, how could it be used? Are there other practices, conventions, or approaches we could follow to address this problem?
The text was updated successfully, but these errors were encountered:
We would like to be able to generalize models and organize them in sets:
Suggestion: support "tagging" models in the cache with a set of searchable string values, using a metadata index.
Caches could then be searched for models with certain tags. Searching should allow for comparisons/regexes so some tags can represent versions.
This is similar to how hashtags are used for github topics, tweets, etc. Each model could have zero or more tags.
May also want to consider extending this to support performance/cost metadata (e.g. number of parameters, quantization level, etc). This can also be used to support "friendly" names for models. Tags should be included in the hash to avoid using them for cross-site communication and protecting them from modification.
Assuming tagging was implemented, how could it be used? Are there other practices, conventions, or approaches we could follow to address this problem?
The text was updated successfully, but these errors were encountered: