Open micahjsmith opened 5 years ago
Hey, I was searching through the documentation and I was wondering if there is direct support for MLBlocks Pipelines? As in, give BTB a json with the MLBlocks Pipeline and start tuning. (I didn't find anything on it on the docs)
There is support for this, though you have to write a bit of glue code.
Basically, you load a mlblocks.MLPipeline
from the JSON, create BTB tunable
using Tunable.from_dict
, and then pass the tunable to a BTBSession
.
Full working example at https://github.com/ballet/predict-life-outcomes/blob/master/src/fragile_families/analysis/search.py#L168 - see lines 161-171
If you find this example works for your use case, please consider refactoring it into a general-purpose code snippet and making a PR to update the BTB docs!
@micahjsmith Thanks! I found a similar solution, though the one you presented does look a bit cleaner. I'll look into updating mine on the next couple of days. Thanks for the help!
Create the following documentation pages:
Reorganize tree, see MLBlocks.
See also: https://hdi-project.github.io/MLBlocks/
See also: #52