Model loading and warm-up takes longer than 60 seconds - Deployment Fails

Im running my own container with xgboost stacked on top of a tensorflow-based network. When I deploy the model the heath check fails, I guess because it takes to long to load the model from S3. If i train xgboost with less rounds, the deployment works but the predictive performance is of course bad.

**So is there a way to load models that take more than 60 seconds to load?** Or are there any other best practices around that? Maybe I got something wrong, because it seems to be a strange restriction...

I'm very thankful for help.

Start by reviewing the deployment health-check behavior, model-loading path, and S3 setup described in the issue. Verify whether SageMaker supports startup times beyond 60 seconds and identify the applicable configuration or best-practice guidance; done means a documented, confirmed answer.

Written by the indexing model from the issue text.