MLflow Triton Plugin

MLflow is a popular open source platform to streamline machine learning development including tracking experiments, packaging code into reproducible runs, and sharing and deploying models. The MLflow Triton plugin is for deploying your models from MLflow to Triton Inference Server. Scripts are included for publishing models, which are in Triton recognized structure, to your MLflow Model Registry.

Supported flavors

MLflow Triton plugin currently supports the following flavors, you may substitute the flavor specification in the example below according to the model to be deployed.

• onnx
• triton

Requirements

• MLflow
• Triton Python HTTP client
• Triton Inference Server

Installation

Pre-built Docker image

Pre-built MLflow Triton Plugin Docker images can be downloaded from NGC.

docker pull nvcr.io/nvidia/morpheus/mlflow-triton-plugin:latest

From source

The plugin can also be installed from the Triton GitHub source using the following commands:

python setup.py install

Quick Start

In this documentation, we will use the files in the Triton Github examples to showcase how the plugin interacts with Triton Inference Server. The onnx_float32_int32_int32 model in examples is a simple model that takes two float32 inputs, INPUT0 and INPUT1, with shape [-1, 16], and produces two int32 outputs, OUTPUT0 and OUTPUT1, where OUTPUT0 is the element-wise summation of INPUT0 and INPUT1 and OUTPUT1 is the element-wise subtraction of INPUT0 and INPUT1.

Start Triton Inference Server in EXPLICIT mode

The MLflow Triton plugin must work with a running Triton server, see documentation of Triton Inference Server for how to start the server. Note that the server should be run in EXPLICIT mode (--model-control-mode=explicit) to exploit the deployment feature of the plugin.

Once the server has started, the following environment variables must be set so that the plugin can interact with the server properly:

• MLFLOW_TRACKING_URI: The URI of the tracking database (default is a SQLite file)
• TRITON_URL: The address to the Triton HTTP endpoint (do not include the URL scheme)
• TRITON_MODEL_REPO: The path to the Triton model repository

Publish models to MLflow

ONNX flavor

The MLflow ONNX built-in functionalities can be used to publish onnx flavor models to MLflow directly, and the MLflow Triton plugin will prepare the model to the format expected by Triton. You may also log config.pbtxt as additonal artifact which Triton will be used to serve the model. Otherwise, the server should be run with auto-complete feature enabled (--strict-model-config=false) to generate the model configuration.

import mlflow.onnx
import onnx
model = onnx.load("examples/onnx_float32_int32_int32/1/model.onnx")
mlflow.onnx.log_model(model, "triton", registered_model_name="onnx_float32_int32_int32")

Triton flavor

For other model frameworks that Triton supports but not yet recognized by the MLflow Triton plugin, the publish_model_to_mlflow.py script can be used to publish triton flavor models to MLflow. A triton flavor model is a directory containing the model files following the model layout. Below is an example usage:

python publish_model_to_mlflow.py --model_name onnx_float32_int32_int32 --model_directory <path-to-the-examples-directory>/onnx_float32_int32_int32 --flavor triton

Deploy models tracked in MLflow to Triton

Once a model is published and tracked in MLflow, it can be deployed to Triton via MLflow's deployments command, the following command will download the model to Triton's model repository and request Triton to load the model.

mlflow deployments create -t triton --flavor triton --name onnx_float32_int32_int32 -m models:/onnx_float32_int32_int32/1

Perform inference

After the model is deployed, the following command is the CLI usage to send inference request to a deployment.

mlflow deployments predict -t triton --name onnx_float32_int32_int32 --input-path <path-to-the-examples-directory>/input.json --output-path output.json

The inference result will be written in output.json and you may compare it with the results in expected_output.json

MLflow Deployments

"MLflow Deployments" is a set of MLflow APIs for deploying MLflow models to custom serving tools. The MLflow Triton plugin implements the following deployment functions to support the interaction with Triton server in MLflow.

Create Deployment

MLflow deployments create API deploys a model to the Triton target, which will download the model to Triton's model repository and request Triton to load the model.

To create a MLflow deployment using CLI:

mlflow deployments create -t triton --flavor triton --name model_name -m models:/model_name/1

To create a MLflow deployment using Python API:

from mlflow.deployments import get_deploy_client
client = get_deploy_client('triton')
client.create_deployment("model_name", "models:/model_name/1", flavor="triton")

Delete Deployment

MLflow deployments delete API removes an existing deployment from the Triton target, which will remove the model in Triton's model repository and request Triton to unload the model.

To delete a MLflow deployment using CLI

mlflow deployments delete -t triton --name model_name

To delete a MLflow deployment using CLI

from mlflow.deployments import get_deploy_client
client = get_deploy_client('triton')
client.delete_deployment("model_name")

Update Deployment

MLflow deployments update API updates an existing deployment with another model (version) tracked in MLflow, which will overwrite the model in Triton's model repository and request Triton to reload the model.

To update a MLflow deployment using CLI

mlflow deployments update -t triton --flavor triton --name model_name -m models:/model_name/2

To update a MLflow deployment using Python API

from mlflow.deployments import get_deploy_client
client = get_deploy_client('triton')
client.update_deployment("model_name", "models:/model_name/2", flavor="triton")

List Deployments

MLflow deployments list API lists all existing deployments in Triton target.

To list all MLflow deployments using CLI

mlflow deployments list -t triton

To list all MLflow deployments using Python API

from mlflow.deployments import get_deploy_client
client = get_deploy_client('triton')
client.list_deployments()

Get Deployment

MLflow deployments get API returns information regarding a specific deployments in Triton target.

To list a specific MLflow deployment using CLI

mlflow deployments get -t triton --name model_name

To list a specific MLflow deployment using Python API

from mlflow.deployments import get_deploy_client
client = get_deploy_client('triton')
client.get_deployment("model_name")

Run Inference on Deployments

MLflow deployments predict API runs inference by preparing and sending the request to Triton and returns the Triton response.

To run inference using CLI

mlflow deployments predict -t triton --name model_name --input-path input_file --output-path output_file

To run inference using Python API

from mlflow.deployments import get_deploy_client
client = get_deploy_client('triton')
client.predict("model_name", inputs)

Container Security

NVIDIA has observed false positive identification, by automated vulnerability scanning tools, of packages against National Vulnerability Database (NVD) security bulletins and GitHub Security Advisories (GHSA). This can happen due to package name collisions (e.g., Mamba Boa with GPG Boa, python docker SDK with docker core). NVIDIA is committed to providing the highest quality software distribution to our customers. The containers are purpose built for Morpheus use cases, have several dependencies, and are not intended for general purpose utility such as web hosting.

In this release, we note the following vulnerabilties:

• GHSA-w3h3-4rj7-4ph4+gunicorn-21.2.0: A CVE discovered late in the release cycle which affects MLflow.
• CVE-2022-37454+pypy-7.3.15: A false positive CVE.

MLflow License

MLflow is licensed under the Apache Software License 2.0.