ga-model-deployment

Version: v2.0.1

Table of contents

1. Getting started
2. Changelog
3. License

Getting started

.env configuration (Required)

As REQUIRED by the script, several ENV variables (which are handled automatically by github in GithubAction cases) needs to be provided. The best option is to create a .env file with the following variables (remove comments):

Warning Please avoid use doubled quotes (e.g "example value") on .env file, simply provide the value right after the = sign. As the script might failed

Considering you have your aws cli installed and configured with storage access AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY does not have to be provided in .env

WORKER_ENV=production
NUTSHELL_MAX_CROP=4000
NUTSHELL_MODE=worker
NUTSHELL_MODEL_SERVING_NAME=exported-barley-protein-v1-22-1-devops
NUTSHELL_WORKER_MODEL_FILE_LOC_ID=1957034
NUTSHELL_WORKER_MODEL_PREDICT_TIMEOUT_S=60
NUTSHELL_MODEL_VERSION=v1.22.1-devops
NUTSHELL_LISTENING_PORT=8080
NUTSHELL_MODEL_PATH=gs://inarix-models/export/XXX
NUTSHELL_MODEL_CONFIG=''
LABEL_TEMPLATE_SLUG=barley_variety_predicted_v7
SHARED_CACHE_PATH=/mnt/filestore
MODEL_TEMPLATE_ID=15
EXPORTED_MODEL_ID=834309eb-e11c-4f2c-82cf-5139072d5eca
LOKI_FILE_LOCATION_ID=1957033
GOOGLE_APPLICATION_CREDENTIALS=/app/client_secrets.json
MODEL_HELM_CHART_VERSION=2.5.1
GITHUB_REPOSITORY=inarix/mt-XX
ARGOCD_ENTRYPOINT=https://argocd.inarix.com/api/v1
ARGOCD_TOKEN=""

Launching Metaflow run on local machine

The main entrypoint is the cli-compatible bookish.py You can simply launch a model deployment scenario with the following syntax: make local

Launching Metaflow run on ArgoWorkflows

As I wanted it to be ran ASAP since this could be used for 🔥🔥 reasons, I used the make argo

Changelog

SEE CHANGELOG IN CHANGELOG.md

Annexes

1. Metaflow documentation

2. Config documentation on code