Ray

Headlines · 4

1. Oct 21
   Databricks AI Runtime reaches general availability

   Databricks announced the general availability of AI Runtime (AIR), a fully managed GPU training layer inside its lakehouse that combines serverless GPU capacity, developer tooling and Unity Catalog governance. Databricks says customers can run multinode training with seconds-level launch latency on RDMA-enabled GPUs, bring their own code, containers and Ray workloads, and debug through a new MLflow observability experience. The company says hundreds of companies, including Robinhood, PicPay, Merck, Flo Health, TinyFish and FactSet, already use AIR in production, though these adoption and performance claims are vendor-supplied and not independently verified.

   Databricks Blog · 🔥 0
2. Oct 20
   AWS fine-tunes GPT-OSS-20B for multilingual reasoning

   AWS Machine Learning Blog reports that supervised fine-tuning raised GPT-OSS-20B’s reasoning-language accuracy from 22% to 100% on a 50-case test covering English, French, German, Spanish and Italian. Applying GRPO after SFT increased answer-language accuracy from 62% to 70%, while a Qwen2.5-7B control experiment added 6 points on multilingual reasoning but produced no gain on GSM8K.

   AWS Machine Learning Blog · 🔥 0
3. Oct 20
   Nscale Joins PyTorch Foundation as Platinum Member

   Nscale has joined the PyTorch Foundation as a Platinum member, gaining a seat on the organization’s governing board. The AI cloud provider said it will continue investing in Ray following its planned acquisition of Anyscale and contribute to open-source infrastructure for distributed AI workloads.

   PyTorch Blog · 🔥 0
4. Oct 12
   AWS shows multi-node inference with Ray LLM Deep Learning Containers

   An AWS Machine Learning Blog post walks through deploying one large language model across two GPU nodes on Amazon EKS using the multi-node Ray LLM Deep Learning Container, with separate prefill and decode nodes behind a single OpenAI-compatible endpoint. The container ships Ray Serve LLM, NVIDIA's NIXL library and the Elastic Fabric Adapter stack, and the example uses Qwen3.5-9B to show prefill/decode disaggregation configured entirely through YAML with no application code. The post also warns that NIXL silently falls back to a TCP-speed transport when it cannot use libfabric, so it recommends checking worker logs to confirm the backend.

   AWS Machine Learning Blog · 🔥 0

Experience and discussion from the community

Share my Ray experienceAsk about Ray