dotnet / dotnet/machinelearning
[AutoML] refactor internal benchmarking so it doesn't need internalsvisible AutoML
Open
area-AutoML
Priority:2
- Dominant language
- C#
- Stars
- 9.4k
- Forks
- 2k
- Avg merge
- 2d 20h
- Merged PRs (30d)
- 11
Description
The reason we use internalsvisible currently is to be able to see what pipleline did an iteration run, including all the details like hyperparam values. It is present in an internal AutoML structure.
Ideal solution to this is if ML.NET starts supporting an interface that allows exploring an istantiated pipeline.
Second best, we can come up with some logging or ToString() solutions that will expose what we need.
Contributor guide
Assessment
This issue has not been assessed yet.