xgboost library 0.9 vs 1.3.3 - xgboost

Question

I worked for some time on hyper-parameters tuning on a XGBRegressor based on xgboost library version 0.9 (Python for Anaconda on Windows).
After installing the latest Anaconda and xgboost(version 1.3.3 this time) on a different PC, I noticed different results on the same code. Here is the simplified code:
from xgboost import XGBRegressor
params = {
'max_depth': 9,
'n_estimators': 400,
'learning_rate': 0.1,
'min_child_weight': 2,
'reg_alpha': 0.05,
'reg_lambda': 1,
'colsample_bytree': .8,
'colsample_bynode': 1,
'colsample_bylevel': .9,
'objective': 'reg:squarederror'
}
model = XGBRegressor(**params)
model.fit(X, Y)
If I am using model to make predictions, I get different results with xgb 1.3.3. I tried to match the hyper parameters by adding all the new default parameters I found in the 1.3.3 library
'base_score': 0.5,
'booster': 'gbtree',
'gamma': 0,
'max_delta_step': 0,
'random_state': 0,
'scale_pos_weight': 1,
'subsample': 1,
'seed': 0
but still the same result. Can you help me adapting the code in order to get the same results on the new environment. I need this to avoid reworking on tuning.
Thank you!

Answer 1

You need to find out which version of XGboost have you used for the desired results. If it xgboost 0.9, you need to install this version. To do so, first uninstall xgboost
pip uninstall xgboost
and then, install xgboost-0.9
pip install xgboost==0.9
You can also look for updates here