fil_backend

fil_backend copied to clipboard

…endency file The issue occurs during fil_backend dockerfile build. We will get the `301` result in actual if trying to request "https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.0-1_all.deb" directly. So the dpkg installer just reports a...

YanShenChun

I deployed multiple models on the Triton server(docker), BERT models are using GPU, and XGBoost models are using CPU. Now I want to limit the number of CPU cores used...

SkyM31

Provide ARM release to allow support for FIL backend on Jetson and other ARM platforms

20

Hello, The [FIL backend installation instructions](https://github.com/triton-inference-server/fil_backend/blob/66543e5ea2710e44ba36f2b4b1d20fcc28eefa04/docs/install.md) indicate that: > The FIL backend is a part of Triton and can be installed via the methods described in the [main Triton documentation](https://github.com/triton-inference-server/server#build-and-deploy)....

blthayer

LLVM ERROR: out of memory

18

root@2ff024ed2346:/opt/tritonserver/tmp/simple-xgboost# python3 sample.py Test Accuracy: 51.24 /usr/local/lib/python3.10/dist-packages/xgboost/core.py:160: UserWarning: [09:16:55] WARNING: /workspace/src/c_api/c_api.cc:1240: Saving into deprecated binary model format, please consider using `json` or `ubj`. Model format will default to JSON in...

sandeepb2013

* Use `use_experimental_optimizations` flag to selectively enable the new FIL * Enable the new FIL for both CPU and GPU inference workload. Note: requires https://github.com/rapidsai/cuml/pull/5559 to function. * Implement common...

hcho3

Rewrite tests

1

This PR moves the model generation and loading into pytest. This achieves a few things: - Generation of configurations and models on the fly by pytest - Testing prediction results...

RAMitchell

[BUG] cuml binary classification models do not observe threshold

1

When setting threshold=0.9, cuml models for CPU or GPU do not appear to observe the threshold. Other model types seem to be working.

RAMitchell

Multiclass models must have the paramers predict_proba=true AND output_class=true in order to predict probabilities, otherwise we get the following backend error: ``` terminate called after throwing an instance of 'raft::exception'...

RAMitchell

[FEA] Provide CatBoost Support

4

riaris

[Bug] Wrong output shapes for Shapley values

3

In the case of a binary classification model from sklearn we expect the output for both positive and negative classes (this would be consistent with the normal prediction output). As...

RAMitchell