Loggers are under-documented and confusing

[Make42]

1. https://clear.ml/docs/latest/docs/references/sdk/logger/ says that loggers are supposed to capture inputs and outputs. BUT, then

:param title: The title (metric) of the table.

suggests that performance measurements are supposed to be capture. This suggests it is not to capture inputs.

2. How do I log configuration (which is input)?
3. How do I log metadata for Datasets?
4. Loggers seem to be very tailored towards neural networks and similarly iteratively improving algorithms. I do not have that - are loggers not for me then? I am confused.
5. I do not understand why the "series" is named as such. In math, a series is the sum of a sequence. I am not aware that "series" is a term in computer science or machine learning. The term "time series" exists, but this is a misnomer and a "time series" is a sequence. Do you mean "sequence"?
6. I do not understand why the "scalar" is named as such. In math, a scalar is used in the context of [vector spaces](https://en.wikipedia.org/wiki/Scalar_(mathematics), including the result of an inner product. In programming it sometimes relates to a non-composite value. But, unless I misunderstand, the logger means to log an "element" or "member" of a sequence, right?

Esp. regarding questions 4-6, I get the feeling that I approach this from a certain (math-y) angle, that is not what the designers of the ClearML SDK had in mind and that is why it makes no sense in my head. What is the angle, that I should take to wrap my head around this?