Prophecis Project Overview
Prophecis is an innovative machine learning platform developed by WeBank, designed to streamline the machine learning process from development to deployment. It combines various open-source machine learning frameworks and offers robust capabilities for managing machine learning compute clusters across multiple tenants. Additionally, it provides comprehensive container deployment and management services specifically optimized for production environments.
Architecture
Key Services
Prophecis is composed of five major services:
-
Prophecis Machine Learning Flow: This distributed tool aids in model training, both standalone and distributed, and supports popular machine learning frameworks such as TensorFlow, Python, and XGBoost. It facilitates the entire machine learning pipeline, right from model creation to deployment.
-
Prophecis MLLabis: An online development and exploration environment, MLLabis is based on Jupyter Lab and allows users to conduct machine learning tasks across GPU and Hadoop clusters. It supports programming languages like Python, R, and Julia, and includes essential plugins like Debug and TensorBoard for enhanced development experience.
-
Prophecis Model Factory: This service is dedicated to model storage, deployment, management, and AB testing, streamlining the lifecycle of machine learning models.
-
Prophecis Data Factory: Offering tools for feature engineering, data labeling, and material management, this facet of Prophecis aids in the efficient preparation and handling of data for machine learning tasks.
-
Prophecis Application Factory: Co-developed by Webank's big data platform team and AI Department, this service leverages QingCloud's KubeSphere system to provide CI/CD tools, DevOps capabilities, and GPU cluster monitoring.
Features
-
Complete Machine Learning Lifecycle: Prophecis seamlessly integrates into workflows thanks to its capability to support the entire machine learning process—from data handling and preprocessing to model training, evaluation, and deployment.
-
One-Click Model Deployment: With Prophecis MF, models can be deployed as restful APIs or RPC interfaces swiftly, enabling a seamless integration with business systems.
-
Enterprise-Grade Management: Leveraging community open-source programs, Prophecis delivers a full-fledged enterprise-level platform for application release, monitoring, service management, and log handling, ensuring a smooth deployment in production environments.
Getting Started
For those new to Prophecis, the platform offers:
- A Quick Start Guide to help users begin their journey with Prophecis effectively.
- A Configuration Guide detailing key configuration settings.
Development and Contribution
Developers keen to extend Prophecis's capabilities can refer to the Development Guide for guidance on development processes.
Prophecis's team encourages contributions. Community input is welcomed and highly valued to drive further development and improvement of the platform.
Communication and Support
For immediate assistance or to engage with the community, users are encouraged to raise issues and join the Prophecis communication groups via WeChat and QQ for direct support.
Prophecis is available under the Apache 2.0 license, ensuring it remains accessible and open for widespread use and adaptation.