dmitryikh / leaves

pure Go implementation of prediction part for GBRT (Gradient Boosting Regression Trees) models from popular frameworks
MIT License
432 stars 72 forks source link

predict result different with python #90

Open aibear-huang opened 1 year ago

aibear-huang commented 1 year ago

github.com/dmitryikh/leaves v0.0.0-20230708180554-25d19a787328 predict 57.46251861844212 python xgboost=0.90 predict 10.252338 this predict is correct python xgboost=1.7.2 predict 57.462517 my xgb trained by xgboost=0.90 (leaves load xgboost will get error when version is not 0.90 )

here is my go code: model, err := leaves.XGEnsembleFromFile(modelfile, false) feats := []float64{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 13032.277889490128, 0, 0, 0, 0, 0, 0, 0, 0, 0} ret := model.PredictSingle(feats, 0) fmt.Print(ret)

it there something wrong with my code ? a.txt is my xgb0.9 model file a.txt could someone help ?thank your advance

aibear-huang commented 1 year ago

image i found the cause but don't know how to solve