h2o-mlops 0.62.1a4

Python client for H2O MLOps.

An H2O MLOps Python Client for Regular Folks

This project is a work in progress and the API may change. Please send questions or feedback to [e-mail].

Example

import h2o_mlops
import httpx
import time

First we have to connect to MLOps. In this example the client is detecting credentials and configuration options from the environment.

mlops = h2o_mlops.Client()

Everything Starts with a Project

A project is the main base of operations for most MLOps activities.

project = mlops.projects.create(name="demo")

mlops.projects.list(name="demo")

    | name   | uid
----+--------+--------------------------------------
  0 | demo   | 45e5a888-ec1f-4f9c-85ca-817465344b1f

You can also do project = mlops.projects.get(...).

Upload an Experiment

experiment = project.experiments.create(
    data="/Users/jgranados/Downloads/GBM_model_python_1649367037255_1.zip",
    name="experiment-from-client"
)

Some experiment attributes of interest.

experiment.scoring_artifact_types

['h2o3_mojo']

experiment.uid

'e307aa9f-895f-4b07-9404-b0728d1b7f03'

Existing experiments can be viewed and retrieved.

project.experiments.list()

    | name                   | uid                                  | tags
----+------------------------+--------------------------------------+--------
  0 | experiment-from-client | e307aa9f-895f-4b07-9404-b0728d1b7f03 |

You can also do experiment = projects.experiments.get(...).

Create a Model

model = project.models.create(name="model-from-client")

Existing models can be viewed and retrieved.

project.models.list()

    | name              | uid
----+-------------------+--------------------------------------
  0 | model-from-client | d18a677f-b800-4a4b-8642-0f59e202d225

You can also do model = projects.models.get(...).

Register an Experiment to a Model

In order to deploy a model, it needs to have experiments registered to it.

model.register(experiment=experiment)

model.versions()

    |   version | experiment_uid
----+-----------+--------------------------------------
  0 |         1 | e307aa9f-895f-4b07-9404-b0728d1b7f03

model.get_experiment(model_version="latest").name

'experiment-from-client'

Deployment

What is needed for a single model deployment?

• project
• model
• environment
• scoring runtime
• name for deployment

We already have a project and model. Let us look at how to get the environment.

project.environments.list()

    | name   | uid
----+--------+--------------------------------------
  0 | DEV    | a6af758e-4a98-4ae2-94bf-1c84e5e5a3ed
  1 | PROD   | f98afa18-91f9-4a97-a031-4924018a8b8f

environment = project.environments.list(name="DEV")[0]

You can also do project.environments.get(...).

Next we'll get the scoring_runtime for our model type. Notice we're using the artifact type from the experiment to filter runtimes.

mlops.runtimes.scoring.list(artifact_type=model.get_experiment().scoring_artifact_types[0])

    | name              | artifact_type   | uid
----+-------------------+-----------------+-------------------
  0 | H2O-3 MOJO scorer | h2o3_mojo       | h2o3_mojo_runtime

scoring_runtime = mlops.runtimes.scoring.list(
    artifact_type=model.get_experiment().scoring_artifact_types[0]
)[0]

Now we can create a deployment.

deployment = environment.deployments.create_single(
    name = "deployment-from-client",
    model = model,
    scoring_runtime = scoring_runtime
)

while not deployment.is_healthy():
    deployment.raise_for_failure()
    time.sleep(5)
    
deployment.status()

'HEALTHY'

Score

Once you have a deployment, you can score with it through the HTTP protocol.

response = httpx.post( 
    url=deployment.url_for_scoring,
    json=deployment.get_sample_request()
)

response.json()

{'fields': ['C11.0', 'C11.1'],
 'id': 'e307aa9f-895f-4b07-9404-b0728d1b7f03',
 'score': [['0.49786656666743145', '0.5021334333325685']]}

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.62.1a4] - 2023-12-01

Added

• Owner attribute for deployments, experiments, models, and projects.
• Ability to view deployment Kubernetes options (including requests, limits, affinity, and toleration).
• Ability to update deployment Kubernetes options (including scaling deployment down to zero resource usage).
• Ability to view deployment security options.
• Ability to update deployment security options (including changing passphrase for existing deployments).
• Ability to enable/disable monitoring for new and existing deployments.

Changed

• "UNHEALTY" status typo corrected to "UNHEALTHY".

[0.62.1a3] - 2023-11-20

Added

• Support for experiment artifacts.

Changed

• experiment.artifact_types renamed to experiments.scoring_artifact_types.

Fixed

• List methods not returning over 100 entries.

[0.62.1a2] - 2023-11-13

Added

• Support for experiment comments.
• Support for experiment tags.
• Integration of https://github.com/h2oai/cloud-discovery-py.

[0.62.1a1] - 2023-10-16

Changed

• Use MLOps 0.62.1 backend.

[0.61.1a3] - 2023-07-28

Changed

• MLOpsClient class renamed to Client.
• _mlops_backend attribute renamed to _backend.