How to show progress of sklearn.multioutput.MultiOutputRegressor and XGBRegressor?

Question

Is it possible to show the training progress of the MultiOutputRegressor in sklearn? When a huge dataset is processed, my program runs a long time and I have no clue how long it will take. I have shortened my program to a minimal working example below.

import numpy as np
from sklearn.multioutput import MultiOutputRegressor
import xgboost as xgb

df = np.arange(50).reshape(10,5)
train = df[:8]
test = df[8:]
X_train = train[:,0:-2]
X_test  = test[:,0:-2]
y_train = train[:,-2:]
y_test  = test[:,-2:] 

eval_set = [(X_test, y_test)]
multioutputregressor = MultiOutputRegressor(xgb.XGBRegressor(eval_set=eval_set, verbose_eval=True))
multioutputregressor.fit(X_train, y_train)
predictions = multioutputregressor.predict(X_test)
print(predictions)

Answer 1

You can consider modifying the code of MultiOutputEstimator or XGBModel to introduce some debugging output. These seem to be the classes that implement the fitting logic of the two libraries. The corresponding source files should also be provided in the installations on your local hard drive.

For example, you can print information of when the separate threads are starting and stopping in MultiOutputEstimator.fit (inherited and thus reused in MultiOutputRegressor), lines 167-169. Also, consider the tip in the documentation of the n_jobs parameter:

If 1 is given, no parallel computing code is used at all, which is useful for debugging.

You can use a similar approach to expand the debugging in XGBModel.fit (the basis of XGBRegressor).