Update tuner to allow single value parameters#1686
Conversation
Codecov Report
@@ Coverage Diff @@
## main #1686 +/- ##
=========================================
+ Coverage 100.0% 100.0% +0.1%
=========================================
Files 240 240
Lines 18652 18658 +6
=========================================
+ Hits 18644 18650 +6
Misses 8 8
Continue to review full report at Codecov.
|
| tuner.add({'Mock Classifier': {'param a': 0, 'param b': 1.0, 'param c': 'option a'}}, np.inf) | ||
| tuner.add({'Mock Classifier': {'param a': 0, 'param b': 1.0, 'param c': 'option a'}}, None) | ||
| tuner.propose() | ||
| print(random_state) |
There was a problem hiding this comment.
remove unnecessary print statements
|
It seems you have a exception in grid_search_tuner that is uncovered by some of the tests you dropped. It's only one line but you can find it in the codecov.io coverage changes. |
chukarsten
left a comment
chukarsten
left a comment
There was a problem hiding this comment.
I think you might just need to address the uncovered exception and this is gucci.
jeremyliweishih
left a comment
jeremyliweishih
left a comment
There was a problem hiding this comment.
Looks great to me but could you add coverage making sure that the tuners can propose with single values as well?
3 participants
first step to solve #1625
We want to allow tuners to accept single values as input; however, we will not add these values to the search space. These values are then treated as fixed values. We use this to help with the implementation for #1625