Can't load xgboost models created with Sagemaker Estimator


Stumbled upon this while trying to evaluate my xgboost model:

model = pickle.load(open("./data/xgboost-model", "rb"))
UnpicklingError: unpickling stack underflow

The model was trained using :

container = image_uris.retrieve(framework='xgboost', region=boto3.Session().region_name, version='1.3-1')

Any ideas on how I can load the model? Thanks.

  • Hi Dani, I'm able to load the XGBoost model trained using a built-in container using the same script you've used above. Can you re-download the pickle file from S3 and try it again? From the error, it looks like it might be an issue with partial file download.

asked 2 years ago970 views
1 Answer


I see that you are using XGBoost version = 1.3-1

Looks like the latest xgboost have changed the way the model is saved or loaded.

If you are using pickle to load the model, then try using the previous version of xgboost, such as 1.2-1, which will get loaded successfully.

If you would like to use version 1.3-1, you can load it the following way -


import xgboost as xgb

model = xgb.Booster()



If you still continue to face the issue, I would recommend you to open a case with AWS Premium Support SageMaker team so that we can discuss more on the use case along with the implementation required for your use case Open a support case with AWS using the link:

answered 2 years ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions