🧑‍🚀Fiat Copilot

- Fiat Copilot offers a range of tools to assist users in creating ML workflows and smoothing the AI development process from start to finish.

Overview

Fiat Copilot is built with various utilities to help ML developers to design their model production pipeline and model serving applications.

Fiat-Copilot
├── examples
│   ├── assets
│   ├── serving
│   └── storage
├── serving
│   ├── __init__.py
│   ├── attendant.py
│   ├── context.py
│   └── domain.py
├── trainer
│   ├── __init__.py
│   ├── hf_util.py
│   ├── torch_util.py
│   └── xgboost_util.py
├── utils
│   ├── __init__.py
│   ├── alioss.py
│   ├── config.py
│   ├── huaweiobs.py
│   └── tencentcos.py
└── workflow
    ├── __init__.py
    ├── annotations.py
    ├── ray_utils.py
    └── storage.py

This tool utilizes Dagster workflow components and Ray remote task & job components. The Dagster workflow component is a versatile and customizable feature that enables users to create workflows specific to their needs.

Meanwhile, Ray provides primitives that are designed for effectively parallelizing AI and Python applications on a local device and scaling to a cloud or on-premises cluster without requiring any code changes. With Ray AI Runtime (AIR), the Fiat can handle various compute-intensive ML workloads. In addition, Ray can automatically scale in and out during runtime and transparently parallelize tasks and jobs.

Module Details

🌻Workflow🏇Trainer & Predictor💂Serving