Skip to content

Latest commit

 

History

History
92 lines (54 loc) · 3.46 KB

README.md

File metadata and controls

92 lines (54 loc) · 3.46 KB

MLOps Task Server with Prefect

We aim to design a task server that can orchestrate variety of data processing and ML related tasks of priority ranking etc .. -

  1. Encoding
  2. PDF Processing
  3. Clustering
  4. ML specific - Training(Finetuning)/Inference

prefect

Prefect Architecture

Prefect Server is an open source backend that makes it easy to monitor and execute your Prefect flows. Under the hood, Prefect Server is actually a diverse collection of services that provide a persistent record of your runs, current state, and allow for asynchronous scheduling and notifications. It was designed to expose many of the popular developer features of Prefect Cloud in a way that allows for community contributions, collaborations and customisations. In particular, Prefect Server ships out-of-the-box with:

Prefect Server is composed of a few different services:

  • UI: the user interface that provides a visual dashboard for mutating and querying metadata
  • Apollo: the main endpoint for interacting with the server
  • PostgreSQL: the database persistence layer where metadata is stored
  • Hasura: the GraphQL API that layers on top of Postgres for querying metadata
  • GraphQL: the server's business logic that exposes GraphQL mutations
  • Towel: runs utilities that are responsible for server maintenance
  • Scheduler: schedules and creates new flow runs
  • Zombie Killer: marks task runs as failed if they fail to heartbeat
  • Lazarus: reschedules flow runs that maintain an unusual state for a period of time

Users and Agents only need access to the Apollo endpoint, all other services may reside behind a firewall.

Getting started

❯ make help
Available rules:

clean               Delete all compiled Python files 
install             Install dependencies in pyproject.toml (poetry.lock) 
lint                Lint using flake8 
update              Update dependencies in pyproject.toml 

Getting Started with Prefect Core

Agent Types

Prefect supports several different agent types for deploying on different platforms.

See their respective documentation for more information on each type.

This will start Prefect Core with a Docker agent on http://localhost:8080/default.

❯ prefect server start  
❯ prefect agent docker start --show-flow-logs
❯ prefect create project "test-project"

References:

Getting Started with KubernetesFlow on AKS with Prefect Cloud

Create a API key in Prefect Cloud -

❯ prefect backend cloud  
❯ prefect auth login --key $PREFECT_CLOUD_API_KEY   
❯ prefect create project "test-project"