Gpu articles

Request GPUs by capability with Kubernetes DRA

Use Dynamic Resource Allocation to describe the accelerator a workload needs, understand device claims, and plan the integration with Polyaxon.

Trace a failed training run to an unhealthy GPU

Connect Kubernetes DRA allocations, device health, and application failures without confusing missing telemetry with healthy hardware.

GPU sharing on Kubernetes: MIG vs. time-slicing

Compare NVIDIA MIG and GPU time-slicing on Kubernetes, including memory isolation, resource names, scheduling behavior, and workload fit.

GPU jobs stuck Pending on Kubernetes: a debugging guide

Diagnose Pending GPU jobs by checking queue admission, scheduler events, advertised GPU resources, placement constraints, storage, and node capacity.

Prepare an AI platform for next-generation GPUs

Make AI platforms ready for new accelerator generations through portable workload contracts, device-aware scheduling, topology, storage, compatibility, and migration evidence.

Validate GPU networking with NCCL and RCCL

Run collective communication checks as tracked MPI jobs before using a multi-node GPU pool for training.