Deploy and Use a Multiframework Distributed Deep Learning Platform on...
summary
Fabric for Deep Learning (FfDL) is a Kubernetes microservice platform that runs training in containerized learners, stores data and models in object storage, and integrates with Watson Studio and AI Fairness 360. The claim is scalable, resilient distributed training; a deployment on IBM Cloud reaches roughly 3,000 pods. Demos show an automated robustness pipeline that runs adversarial attacks (fast gradient sign) and reports model accuracy drops of over 60%, and the fairness toolkit can reweight or preprocess to reduce bias.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 2:33AI and cloud synergy
- 3:59Fabric for Deep Learning
- 7:07Fabric architecture
- 9:40Demo and robustness checks
- 12:07Submitting training job
- 14:17Robustness checks and deployment
- 17:31AI Fairness 360 demo
- 21:09Q&A
- 29:04Closing
related talks
34mLFN AI Taskforce: Platform-Driven Fede...
38mThe Lifecycle of Artificial Intelligence with IBM's Deep Learning as a Service
21mWhat, Why and How of Federated Machine Learning – Implementation Using...
39mOperating Deep Learning Pipelines Anywhere Using Kubeflow
31mBuilding and Managing a Centralized ML Platform with Kubeflow at...
22mProductive Machine Learning and Deep Learning Projects
42mDeep learning beyond the learning
22mOperationalising Data Science on Cloud Foundry