Error when training and deploying xgboost in script mode

Hi,

For a project, I want to use sagemaker xgboost to predict movie box office.
I want to use my own script for training the model and predicting.
The training data is stored in s3 as csv file.

I tried to implement my own train function by applying the example code:
https://github.com/makexu93/box-office-predictor/blob/main/train_xgboost.py

The error is:
_unexpectedstatusexception: error for training job sagemaker-xgboost-2020-11-06-19-06-38-225: failed. reason: algorithmerror: executeuserscripterror: command "/miniconda3/bin/python -m train_xgboost --eta 0.2 --gamma 4 --max_depth 5 --min_child_weight 6 --num_round 50 --objective reg:linear --silent 0 --subsample 0.7"_

Also, the predict function wont work:
https://github.com/makexu93/box-office-predictor/blob/main/predict.py

The error is:
_unexpectedstatusexception: error hosting endpoint sagemaker-xgboost-2020-11-06-10-03-24-410: failed. reason: the primary container for production variant alltraffic did not pass the ping health check. please check cloudwatch logs for this endpoint.._

Thanks for your help!

Marco

Start by reading the linked train_xgboost.py and predict.py files, then inspect the reported training-job and endpoint health-check errors in the SageMaker logs. Determine why script-mode training fails and why deployment does not pass its ping check; done means both training and prediction work successfully.

Written by the indexing model from the issue text.