Unable to "set_base_margin" and "predict with model's best_ntree_limit" using ONNX runtime to do prediction on XGBoost model

Question

I have converted a XGBoost model into an ONNX model using onnxmltools. However, when I was trying to perform prediction using ONNX runtime, I have 2 issues.

1. In XGBoost API, DMatrix has a "set_base_margin" function, which initializes the output with "base_margin" values. In ONNX runtime, I can't find a function that can achieve the same functionality. ONNX runtime does allow passing in List of NamedONNXValue instances to receive the prediction output, but all pre-populated values in that list are ignored.
2. In XGBoost API, "predict" function has a "ntree_limit" parameter, which doesn't exist in ONNX runtime.

Any help will be appreciated.