[mlflow-triton-plugin] wrong "backend" value when deploying in onnx flavor

[Rusteam]

Description
When deploying an MLflow model with ONNX flavor, backend equals to onnx in config.pbtxt, while it should be backend=onnxruntime or platform=onnxruntime_onnx.
source

Triton Information

• docker image: nvcr.io/nvidia/tritonserver:21.12-py3
• mlflow-triton-plugin=0.1.0

To Reproduce

mlflow deployments create -t triton --flavor onnx --name <model-name> -m <model-uri>

Error source mlflow_triton/deployments.py:

config = '''
backend: "onnx"
default_model_filename: "{}"
'''.format(model_file)

Expected behavior
backend: "onnxruntime"

I can submit a PR!

[Rusteam]

I see it has been fixed here as well