Loading…
Topic
Ray
4 posts about Ray. Every summary links to the original.
- 4 posts
- 3 companies
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
Posts
High-Throughput Structure Prediction with BioNeMo Inference Runtime
Proteome-scale biomolecular structure prediction requires moving large worklists efficiently, and NVIDIA BioNeMo Inference Runtime (BioIR) accelerates supported models on NVIDIA GPUs while retaining a PyTorch workflow. Its end-to-end processor passes InputRequest objects through parsing, tokenization, feature generation, GPU inference, and PDB or mmCIF writing; a Ray backend places complete model replicas on GPUs for concurrent independent inputs, while direct PyTorch integration supports custom code. In a matched 1,000-target human-dimer benchmark, BioIR completed all targets at 58.5K successfully folded residues per allocated GPU-hour versus 20.2K for the tested public implementation, a 2.90× improvement, which ran out of memory on 29 targets. Rated-power-equivalent estimates for one million comparable targets were 11 versus 35 MWh at 8-GPU TDP and 21 versus 64 MWh at full-node maximum power, excluding data-center overhead and several pipeline costs.
Run Ray on TPU, Part 2: Ray AI libraries
Running Ray on TPUs requires multi-host workloads to remain on one intact TPU slice whose hosts share a high-speed ICI link. This guide presents Ray's AI libraries on top of GKE and Ray Core: Ray Serve for serving, Ray Data's iter_jax_batches() for input pipelines and batch inference, and Ray Train's JaxTrainer for distributed JAX training. Serve uses accelerator_config.topology to defer slice placement to the replica, while iter_jax_batches() returns device-sharded JAX arrays and handles ragged final batches through drop, pad, or raise choices. JaxTrainer takes a training function and slice topology, launches one worker per host, and provides checkpointing, fault-tolerant restarts, and multi-slice scale-out; official TPU images and dashboard metrics round out the workflow.
Run Ray on TPU, Part 1: The foundations
Ray 2.55 makes Google Cloud TPUs a first-class accelerator in Ray, with official APIs, pre-built images, and core-library support rather than the earlier experimental path. The central constraint is TPU slicing: a multi-host workload must land on one intact slice because its hosts communicate through the slice’s ICI (Inter-Chip Interconnect). On GKE, the Ray Operator installs KubeRay and a TPU webhook; the webhook labels hosts, while Ray Core uses those labels to reserve a whole slice atomically through the public slice_placement_group() API, with topology and accelerator version as inputs. The guide shows node selection by TPU generation and topology, a multi-host numOfHosts field, and a RayCluster manifest that brings up one worker per host. Ray AI libraries such as Data, Train, and Serve generally invoke this placement primitive themselves, while direct use targets custom distributed workloads; the API is public but alpha.
Two years of vector search at Notion: 10x scale, 1/10th cost
Notion describes two years of scaling vector search for AI Q&A, where semantic embeddings retrieve workspace and connected-tool content by meaning rather than exact keywords. The initial system combined Apache Spark batch indexing with Kafka-driven real-time updates, workspace-based sharding, and generation routing; rapid demand drove a 600x increase in daily onboarding capacity and cleared a waitlist of millions of workspaces by April 2024. A serverless migration cut vector database costs 50 percent from peak usage, while a later move to turbopuffer reduced search-engine spend 60 percent, lowered AWS EMR costs 35 percent, and improved p50 query latency from 70–100ms to 50–70ms. Further work uses xxHash-based page-state caching in DynamoDB to reprocess only changed spans and shifts embeddings generation from Spark and query embeddings away from a third-party API to Ray on Anyscale, with early rollout results pointing to a projected 90+ percent reduction in embeddings infrastructure costs.
Related topics
All topics →Weekly digest
The best engineering reads, once a week.