help needed for xgboost bring-your-own with distributed training

hi guys,
we are trying to create our own Docker image for xgboost distributed training. We have managed to successfully create an image for single instance training, but are unable to figure out how to hook to the xgboost distributed backend.

I see there is an equivalent example for other algos - https://github.com/awslabs/amazon-sagemaker-examples/blob/0e17c31a69ad014bee71fa9c4d700d35cd30a421/advanced_functionality/fairseq_translation/fairseq/train#L112-L119

is there any way that you guys could share an example of a BYO distributed training docker image for xgboost ?

Start with the linked fairseq train example, especially lines 112-119, and compare it with the repository's available SageMaker training examples. Determine what a bring-your-own XGBoost distributed-training Docker example would need to demonstrate; done would be a working example or documented pattern, but the issue does not name an existing XGBoost file or test.

Written by the indexing model from the issue text.