SkyPilot v0.14.0 ships dynamic job groups, multi-user Slurm GA and AWS Trainium support
skypilot_org · x · 2026-10-15
SkyPilot released v0.14.0 with several notable features for multi-cloud compute orchestration:
- Dynamic Job Groups: a running job group can launch parallel eval jobs on the fly, tracked and cancelled together without blocking training
- Multi-User Slurm is now GA: jobs submit as the authenticated user with per-user QoS/account settings and container cluster start/stop
- AWS Trainium/Inferentia (Neuron) schedulable on Kubernetes just like GPUs
- Read-only workspaces for sharing clusters, jobs and logs with non-members
- Conda-free default image
Install: uv pip install "skypilot[all]==0.14.0"
More from Infra
- AMD MI355X beats B200 on DeepSeek V4 Pro per-MW throughput but trails badly on Kimi K3 — ArtificialAnlys · 2026-10-15
- Inworld opens inference service; Flex tier can reject requests when busy at no charge — dr_cintas · 2026-10-15
- 34M-param Nerve-1 Runs on CPUs to Route AI Tasks, Claiming 30% Inference Cost Cut — saheedniyi_02 · 2026-10-15
- Meta's crawler hit 15x Googlebot volume in internet-scale incident, now fixed via Vercel — cramforce · 2026-10-15
- Repeated-Data Pretraining Study Finds Smaller Batches Save GPU-Time Under Fixed Deadlines — jasondeanlee · 2026-10-15
- Trackio gets 2.9x faster launches and 7.6x faster run selection in latest update — Gradio · 2026-10-15