Xgboost regressor converted into Sole model #56
Conversation
|
A couple of errors in tests: one is due to a missing deps, the other one is due to missing Natops dataset on http. Sometimes happen, let's check it in a couple of hours. |
| @@ -2,7 +2,7 @@ name: CI | |||
| on: | |||
| push: | |||
| branches: | |||
| - main | |||
There was a problem hiding this comment.
Shouldn't this change be reverted before merging to main?
There was a problem hiding this comment.
I make always a mess with ci sources: usually I change it after merging into main. Bad habit? I guess so.
| featurenames=nothing, | ||
| keep_condensed=false, | ||
| use_float32::Bool=true, | ||
| kwargs... | ||
| ) | ||
| keep_condensed && error("Cannot keep condensed XGBoost.Node.") | ||
|
|
||
| nclasses = length(classlabels) | ||
| classlabels === nothing || (nclasses = length(classlabels)) |
There was a problem hiding this comment.
isnothing(classlabels) is preferred
There was a problem hiding this comment.
I tried that because I saw in MLJ. I wondered why they had used === nothing instead of isnothing.
Surprisingly, checking with @btime, I discovered that sometimes it's faster than isnothing, but in very few cases, so, I agree with you.
Xgboost regression algo was still missing in XGBoost external package.
Now it's fully developed and tested.