THE COLLECTION

EVERY SIGNAL, SORTED.

Search the whole shelf by topic, tool, term, or plain old curiosity.

SHOWING 2 OF 638 STORIES
ISSUE 541MORNING

PyTorch 2.15, Distributed Training, NCCL, GPU Clusters, Reliability Engineering, Rollback Testing, Flight Recorders, Distributed Checkpoints, CUDA, ROCm and Production Infrastructure

PyTorch 2.15 rewrote the GPU handshake. Test the rollback before the cluster hangs

PyTorch made its rewritten NCCL backend the default for distributed GPU jobs. The release includes hang diagnostics and temporary rollback switches, but the safest migration begins with a canary that proves both paths.

#pytorch#distributed-training#reliability
ISSUE 326MORNING

AI Hardware, PyTorch, Accelerator Integration, Open Source Infrastructure, Compatibility Testing, Compilers and Distributed Computing

PyTorch is building a common entrance exam for new AI chips

New accelerators have spent years proving compatibility through private patches and heroic debugging. PyTorch is turning that work into shared tests, reference backends and a public admission process, which is useful evidence but not a performance diploma.

#ai-hardware#pytorch#open-infrastructure