End-to-End Big Data AI Pipeline on Ray and Apache Spark using Analytics Zoo
Speakers
Jason Dai (Intel), Jason Liu (Inspur), and Yang Wang (Intel)
Schedule
10AM-12PM (East Time), June 19, 2021
| 10:00 - 10:45 | Seamlessly Scaling out Big Data AI on Ray and Apache Spark (slides) | Jason Dai, Intel |
| 10:45 - 11:30 | E2E Smart Transportation CV application in Inspur (using Insight Data-Intelligence platform) (slides) | Jason Liu, Inspur |
| 11:30 - 12:00 | Transparent Acceleration of E2E Computer Vision Pipelines | Yang Wang, Intel |
Description
As AI moves from experimentation to production, a key challenge in practice is how to seamlessly integrate distributed data processing and AI programs into a single unified pipeline. Today, AI researchers and data scientists usually need to go through a mountain of pains to apply AI models to production dataset that is stored in distributed Big Data cluster. Conventional approaches would set up two separate clusters, one for Big Data processing, and the other dedicated to deep learning and AI, with “connector” (or glue code) deployed in between. Unfortunately, this not only introduces a lot of data transfer overhead, but also requires additional efforts for managing separate workflows and systems in production.
To address these challenges, we have open sourced Analytics Zoo, which helps users to productionize their end-to-end Big Data AI pipelines in a distributed fashion. Using the Orca library in Analytics Zoo, users can easily build data-parallel preprocessing (using existing Python libraries such as Pillow or OpenCV for images) for distributed dataset; and then within the same program, the users can directly apply AI models to the preprocessed data (that are partitioned and stored in-memory across large clusters), for transparently distributed training and inference (using TensorFlow, PyTorch, BigDL, OpenVINO, etc.) through simple scikit-learn style APIs.
This tutorial will present the design patterns and tradeoffs for building end-to-end Big Data AI pipelines in production (using Apache Spark and Ray). In addition, it will also share real-world experience and “war stories” of users who have adopted Analytics Zoo to productionize their end-to-end AI pipelines.