fil_backend
fil_backend copied to clipboard
FIL backend for the Triton Inference Server
Hello ! XGBoost recently enabled developers to use categorical features in its models (Nvidia did an article on that : https://developer.nvidia.com/blog/categorical-features-in-xgboost-without-manual-encoding/). From what I understand, we can load a XGBoost...
cuML 24.08 added support for XGBoost UBJSON.
Getting following error when running xgboost model (fil backend) with triton container on MacOS14.3 **Error** ` UNAVAILABLE: Invalid argument: unable to find 'libtriton_fil.so' or 'fil/model.py' for model 'xgboost_model', in /opt/tritonserver/backends/fil`...
Requirement: RapidJSON needs to merge https://github.com/Tencent/rapidjson/pull/2334 first.
Hi Team, I'm working on Autogluon automl framework and I wanted to utilize Triton Inference Server. Does Triton Inference Server support AutoGluon framework? If yes, could you please provide me...
I have an XGBoost model and I want to run it on Triton's FIL backend. My XGBoost model requires support for base_margin, but I couldn't find how to use base_margin...
As seen in https://github.com/triton-inference-server/fil_backend/pull/406, we have lots of version numbers specified in many parts of the CI/CD pipeline. It's easy to leave some components in an outdated state.
I have an XGBoost model that I want to run in Triton server. Currently I'm using the model to get raw prediction probability like this: ```py model = xgb.XGBClassifier() model.load_model(str(model_path))...
https://gitlab-master.nvidia.com/dl/triton/fil_backend/-/jobs/169923674: ``` Mismatched indices: {}""".format( total_tol_desc, mismatch_count, mismatch_proportion * 100, a, b, np.transpose(np.nonzero(diff_mask))) > raise AssertionError(msg) E AssertionError: Arrays have more than 0 mismatched elements. E E Mismatch in 2...