Closed c-w-feldmann closed 1 month ago
Currently a _MolPipeline
object is create dynamically, thus each run it has a new ID. This prevents joblib from loading cache (I suspect)
Edit
It uses hashing, and seems to work. Only downside is, if the Pipeline contains only one Element, or all Elements can be aggregated to one subpipeline, no caching is done.
Currently this feature is not supported, as the cached transformers are not properly set after loading the cached results.
Todos: