Open KEggensperger opened 7 years ago
I don't making the internal benchmark version equal to the version number of the external software is a good idea. This wouldn't allow us to fix bugs/change minor things unless the external software changes.
Besides that, yes, a great idea!
How should we handle different versions of our benchmarks?
For example, if we change a configurationspace, users should be able to see that this space is different from something that was previously used.
I propose to add "version" to the metadictionary. For benchmarks such as autosklearn we use the version of that package, for other benchmarks we manually maintain the version number?