Skip to content

Add support for XGBoost eval_metrics and objective #1103

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Oct 20, 2023
Merged

Add support for XGBoost eval_metrics and objective #1103

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
0bcdfa7
fix column order when label is not last
montanalow Oct 19, 2023
03535ff
use standard name
montanalow Oct 19, 2023
b319615
check for nulls before training
montanalow Oct 19, 2023
4d7285d
support objectives and eval metrics
montanalow Oct 19, 2023
d5e01f9
fmt
montanalow Oct 19, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
use standard name
  • Loading branch information
@montanalow
montanalow committed Oct 19, 2023
commit 03535ff9371abfec113d38eb54b449b95ba5f79d
8 changes: 4 additions & 4 deletions pgml-extension/src/orm/snapshot.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,7 @@ pub(crate) enum Encode {
#[default]
native,
// Encode each category as the mean of the target
target_mean,
target,
// Encode each category as one boolean column per category
one_hot,
// Encode each category as ascending integer values
Expand Down Expand Up @@ -230,7 +230,7 @@ impl Column {
target: &ndarray::ArrayView<f32, ndarray::Ix1>,
) {
// target encode if necessary before analyzing
if self.preprocessor.encode == Encode::target_mean {
if self.preprocessor.encode == Encode::target {
let categories = self.statistics.categories.as_mut().unwrap();
let mut sums = vec![0_f32; categories.len() + 1];
Zip::from(array).and(target).for_each(|&value, &target| {
Expand Down Expand Up @@ -535,7 +535,7 @@ impl Snapshot {
Some(preprocessor) => {
let preprocessor = preprocessor.clone();
if Column::categorical_type(&pg_type) {
if preprocessor.impute == Impute::mean && preprocessor.encode != Encode::target_mean {
if preprocessor.impute == Impute::mean && preprocessor.encode != Encode::target {
error!("Error initializing preprocessor for column: {:?}.\n\n You can not specify {{\"impute: mean\"}} for a categorical variable unless it is also encoded using `target_mean`, because there is no \"average\" category. `{{\"impute: mode\"}}` is valid alternative, since there is a most common category. Another option would be to encode using target_mean, and then the target mean will be imputed for missing categoricals.", name);
}
} else if preprocessor.encode != Encode::native {
Expand Down Expand Up @@ -1014,7 +1014,7 @@ impl Snapshot {
let value = match key.as_str() {
NULL_CATEGORY_KEY => 0_f32, // NULL values are always Category 0
_ => match &column.preprocessor.encode {
Encode::target_mean | Encode::native | Encode::one_hot { .. } => len as f32,
Encode::target | Encode::native | Encode::one_hot { .. } => len as f32,
Encode::ordinal(values) => match values.iter().position(|v| v == key.as_str()) {
Some(i) => (i + 1) as f32,
None => error!("value is not present in ordinal: {:?}. Valid values: {:?}", key, values),
Expand Down