Relative Content

Tag Archive for pythonxgboost

Custom loss function for distribution fitting with XGBoost

I want to train an XGBoost model that takes in 2 (or more) inputs and produces 2 output values. These are to be interpreted as parameters of a probability distribution for the outcome, for example, the mean and log-variance of a Normal distribution. To train XGBoost to output these parameters I want to implement the negative log likelihood (here that of the Normal distribution) as a custom objective function.

Saving XGBoost Python AIP model and features: error no such file or directory

When I follow the instructions in the Python API documentation on how to save the XGBoost model and features, I get the following error:
raise XGBoostError(py_str(_LIB.XGBGetLastError()))
xgboost.core.XGBoostError: [08:28:42] /Users/runner/miniforge3/conda-bld/xgboost-split_1712072580858/work/dmlc-core/src/io/local_filesys.cc:209: Check failed: allow_null: LocalFileSystem::Open “featmap.txt”: No such file or directory