I trained XGBoostRegressor on two versions (2.0.0, 1.7.6) using the same data in python3.8.
# X, y = make_regression(n_samples=num_data, n_features=num_feature, noise=noise, bias=10)
xgbr = XGBRegressor(max_depth=5, n_estimators=129, random_state=0)
xgbr.fit(X, y)
However, the shape of the learned model is very different. Can you explain how the learning method changed with version 2.0.0?