ByteDances Platform for Reinforcement Learning from Human Feedback | Ray Summit 2024 @anyscale
ByteDances Platform for Reinforcement Learning from Human Feedback | Ray Summit 2024  @anyscale
Uploaded October 2024 | Updated September 2026, 2 weeks ago
ByteDance's journey in building a robust video data processing pipeline takes center stage in this Ray Summit 2024 presentation. Haibin Lin, Chi Zhang, and Liguang Xie share their experiences in leveraging Ray's ecosystem to tackle the challenges of creating a video generation model capable of producing realistic scenes from text instructions.

Delving into the technical details, the speakers showcase how they harnessed Ray Core, Ray Data, and Ray Serve to develop a scalable pipeline that efficiently handles massive volumes of video data. The ByteDance team explains the crucial role of Ray's capabilities in dynamic computation scaling and heterogeneous resource orchestration in constructing their complex data processing system.

Rounding out the session, Lin, Zhang, and Xie demonstrate ByteDance's approach to managing Ray infrastructure, along with best practices gleaned from their experience. This talk offers significant value for teams working on AI models that require processing and analysis of extensive multimedia datasets, providing a roadmap for implementing similar large-scale data processing solutions.

--

Interested in more?
- Watch the full Day 1 Keynote: youtu.be/jwZHJthQvXo
- Watch the full Day 2 Keynote youtu.be/Lury2ad6KG8

--

đź”— Connect with us:
- Subscribe to our YouTube channel: youtube.com/@anyscale
- Twitter: https://x.com/anyscalecompute
- LinkedIn: linkedin.com/company/joinanyscale
- Website: anyscale.com
ByteDances Platform for Reinforcement Learning from Human Feedback | Ray Summit 2024Dynamic Scheduling for Large Language Model Serving | Ray Summit 2024How Prime Intellect Builds Scalable Infrastructure for Agentic RL | Ray Summit 2025BentoML or RayServe, You Can Choose Both with BentoRayHow KubeRay Is Evolving for Massive AI Workloads  | Ray Summit 2025RDMA P2P Deep Dive: KvCache Transfer, Weight Updates & MoE Routing at Perplexity | Ray Summit 2025Ray Direct Transport: RDMA Support in Ray Core | Ray Summit 2025Building LLaMA: Metas Director of GenAI Sergey Edunov | Ray Summit 2024Maximizing Compute Efficiency on Anyscale | Ray Summit 2025The PARK Stack: The LAMP Stack of the AI Era | Ben LoricaAgentic Workload Inference at Scale: ByteDance’s AIBrix & DeerFlow | Ray Summit 2025Productionizing LLMs at Scale for Real-World Commerce | Ray on the Road – NYC 2025
Anyscale |

ByteDance's Platform for Reinforcement Learning from Human Feedback | Ray Summit 2024

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER