Azure / fast_retraining

Show how to perform fast retraining with LightGBM in different business cases

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Try gpu_use_dp=false in GPU parameters to see if improves the time

miguelgfierro opened this issue · comments

rounds=200, full dataset, double precission:
"lgbm": {
"performance": {
"AUC": 0.842578378921684,
"Accuracy": 0.7667173145300099,
"F1": 0.7355360689058548,
"Precision": 0.7961293571204864,
"Recall": 0.6835139246819957
},
"test_time": 43.04262933897553,
"train_time": 686.1979219189961
}

gpu full dataset num_rounds=200, single precision:
"lgbm": {
"performance": {
"AUC": 0.8426748987428655,
"Accuracy": 0.7667467751094083,
"F1": 0.7355888744118525,
"Precision": 0.7961184571278634,
"Recall": 0.6836131672978735
},
"test_time": 45.97168072301429,
"train_time": 562.9683960059774
}