kego 0.8.0

Helpers for quick ML analysis of kaggle competitions.

kego

Pipeline

Usage idea:

Run training run by explicitly specifying params:

kego run --model catboost --params learning_rate:0.01 --hp-tune --hp-params max_trees::0:9:log

Run training run from config file.

kego run --config catboostv1

Run training run with overwritten config file params:

kego run --config catboostv1 --params featureset:v1

Following configuration files exist:

# config.yaml
competition:
  - name: ragii
  - deadline: 2026.07.01

featurestore:
  - paths: [kristian@omarchyd, ~/]

models:
  - name: xgboost
    params:
      metric: auc

  - name: resnet
    params:
      n_layers: 10

ensemble:
  method: ridge_stack