# MLOps Framework This is a framework for MLOps that deploys models as functions in Cognite Data Fusion

# User Guide

## Reference guide This assumes you are already familiar with the framework, and acts as a quick reference guide for deploying models using the prediction service, i.e. when model training is performed outside of the MLOps framework. 1. Train model to generate model artifacts 2. Manually upload artifacts to your test environment

• This includes model artifacts generated during training, mapping- and settings-file for the model, scaler object etc. Basically everything that is needed to preprocess the data and make predictions using the trained model.

3. Deploy prediction service to test - This is handled by the CI/CD pipeline in Bitbucket
4. Manually promote model artifacts from test to production
5. Manually trigger deployment of model to production - Trigger in the CI/CD pipeline
6. Call deployed model - See section "Calling a deployed model prediction service hosted in CDF" below

## Getting Started: Follow these steps: - Install package: pip install akerbp.mlops - Set up pipeline file bitbucket-pipelines.yml and config file

mlops_settings.yaml by running this command from your repo's root folder: `bash python -m akerbp.mlops.deployment.setup `

• Fill in user settings and then validate them by running this (from repo root): `python from akerbp.mlops.core.config import validate_user_settings validate_user_settings() ` alternatively, run the setup again: `bash python -m akerbp.mlops.deployment.setup `
• Commit the pipeline and settings files to your repo
• Become familiar with the model template (see folder model_code) and make sure your model follows the same interface and file structure (see [Files and Folders Structure](#files-and-folders-structure))
• Follow or request the Bitbucket setup (described later)

A this point every git push in master branch will trigger a deployment in the test environment. More information about the deployments pipelines is provided later.

## Updating MLOps Follow these steps: - Install a new version using pip, e.g. pip install akerbp.mlops==x, or upgrade your existing version to the latest release by running pip install --upgrade akerbp.mlops - Run this command from your repo's root folder:

`bash python -m akerbp.mlops.deployment.setup ` This will update the bitbucket pipeline with the newest release of akerbp.mlops and validate your settings. Once the settings are validated, commit changes and you're ready to go!

## General Guidelines Users should consider the following general guidelines: - Model artifacts should not be committed to the repo. Folder model_artifact

does store model artifacts for the model defined in model_code, but it is just to help users understand the framework ([see this section](#model-manager) on how to handle model artifacts)

• Follow the recommended file and folder structure ([see this section](#files-and-folders-structure))
• There can be several models in your repo: they need to be registered in the settings, and then they need to have their own model and test files
• Follow the import guidelines ([see this section](#import-guidelines))
• Make sure the prediction service gets access to model artifacts ([see this section](#model-manager))

## Configuration MLOps configuration is stored in mlops_settings.yaml. Example for a project with a single model: ```yaml model_name: model1 human_friendly_model_name: 'My First Model' model_file: model_code/model1.py req_file: model_code/requirements.model artifact_folder: model_artifact artifact_version: 1 # Optional test_file: model_code/test_model1.py platform: cdf dataset: mlops python_version: py39 helper_models:

• my_helper_model

info:

prediction: &desc

description: 'Description prediction service, model1' metadata:

required_input:

• ACS
• RDEP
• DEN

training_wells:

• 3/14
• 2/7-18

input_types:

• int
• float
• string

units:

• s/ft
• 1
• kg/m3

output_curves:

• AC

output_units:

• s/ft

petrel_exposure: False imputed: True num_filler: -999.15 cat_filler: UNKNOWN

owner: data@science.com

training:

<< : *desc description: 'Description training service, model1' metadata:

required_input:

• ACS
• RDEP
• DEN

output_curves:

• AC

hyperparameters:

learning_rate: 1e-3 batch_size: 100 epochs: 10

``` | Field | Description | | --------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | model_name | a suitable name for your model. No spaces or dashes are allowed | | human_friendly_model_name | Name of function (in CDF) | | model_file | model file path relative to the repo's root folder. All required model code should be under the top folder in that path (model_code in the example above). | | req_file | model requirement file. Do not use .txt extension! | | artifact_folder | model artifact folder. It can be the name of an existing local folder (note that it should not be committed to the repo). In that case it will be used in local deployment. It still needs to be uploaded/promoted with the model manager so that it can be used in Test or Prod. If the folder does not exist locally, the framework will try to create that folder and download the artifacts there. Set to null if there is no model artifact. | | artifact_version (optional) | artifact version number to use during deployment. Defaults to the latest version if not specified | | test_file | test file to use. Set to null for no testing before deployment (not recommended). | | platform | deployment platforms, either cdf (Cognite) or local for local testing. | | python_version | If platform is set to cdf, the python_version required by the model to be deployed needs to be specified. Available versions can be found [here](https://cognite-sdk-python.readthedocs-hosted.com/en/latest/functions.html#create-function) | | helper_models | Array of helper models using for feature engineering during preprocessing. During deployment, iterate through this list and check that helper model requirements are the same as the main model. For now we only check for akerbp.mlpet | | dataset | CDF Dataset to use to read/write model artifacts (see [Model Manager](#model-manager)). Set to null is there is no dataset (not recommended). | | info | description, metadata and owner information for the prediction and training services. Training field can be discarded if there's no such service. |

Note:

all paths should be unix style, regardless of the platform.

Notes on metadata:

We need to specify the metadata under info as a dictionary with strings as keys and values, as CDF only allows strings for now. We are also limited to the following - Keys can contain at most 16 characters - Values can contain at most 512 characters - At most 16 key-value pairs - Maximum size of entire metadata field is 512 bytes

If there are multiple models, model configuration should be separated using ---. Example: `yaml model_name: model1 human_friendly_model_name: 'My First Model' model_file: model_code/model1.py (...) --- # <- this separates model1 and model2 :) model_name: model2 human_friendly_model: 'My Second Model' model_file: model_code/model2.py (...) `

## Files and Folders Structure All the model code and files should be under a single folder, e.g. model_code. Required files i