v1.30.0 - Cloud HPC Toolkit A3 VM + NeMo Framework Solution

What's Changed

Key New Features 🎉

• Introduction of the Cloud HPC Toolkit A3 VM family blueprint featuring
  • A Slurm cluster composed of A3 VMs each with 8 NVIDIA H100 GPUs
  • An example for running the NVIDIA NeMo framework
  • An example for running the common nccl-tests benchmark

Module Improvements 🔨

• Add support for startup script per nodeset by @mr0re1 in #2296
• Allocate IP for vm-instance by @lemaitre-aneo in #2219
• Add random hex to batch job id and specify id when submitting by @nick-stroud in #2259
• HTCondor: support user-managed secret replication by @tpdownes in #2340
• Support setting system default web proxy in Windows startup by @tpdownes in #2351

Improvements 🛠

• Add TPU v4 blueprint and tutorial to demonstrate running TPU workload by @harshthakkar01 in #2287
• Update parameters for TPU nodeset module and add precondition checks and bump TPU to v3 by @harshthakkar01 in #2293
• Add Slurm v6 version for image builder blueprint by @harshthakkar01 in #2297
• Allow ghpc deploy blueprint.yaml by @mr0re1 in #2323
• Slurm GCP version update; will cooldown before deleting orphan nodes by @nick-stroud in #2322
• Add SlurmGCP v6 example of slurm compatible with startup scripts and integration test by @harshthakkar01 in #2346

Version Updates ⏫

• Adopt Slurm-GCP release 5.10.6 by @tpdownes in #2352

Bug fixes 🐞

• Added enable_devel for packer build to fix issue with bp by @cdunbar13 in #2334

New Contributors

• @lemaitre-aneo made their first contribution in #2219

Full Changelog: v1.29.0...v1.30.0