A practical, evidence-based framework for collecting and handling InfiniBand LFT, topology, and UFM snapshot data without exposing the management plane.
Read MoreA source-grounded tour of Transformer Engine’s fused-attention dispatch: cuDNN Graph API, versioned backend eligibility, Blackwell paths, kernel-pipelining context, and deterministic training constraints.
Read MoreA technical guide to the architecture, repository layout, parallelism-aware conversion mechanics, and an import→fine-tune→export workflow for NVIDIA’s Megatron stack.
Read MoreA beginner-friendly, code-oriented introduction to Transformer attention, multi-head attention, online softmax, and a tiled CUDA FMHA forward kernel.
Read MoreA practical guide to integrating Enroot, Pyxis, and Slurm SPANK for unprivileged GPU and MPI containers, from architecture and deployment to troubleshooting.
Read More