What's Changed

New Features

• A4X support by @scaliby in #643
• Merge main to develop by @FIoannides in #657
• Feat: Add --skip-validation flag to bypass system dependency checks by @RexBearIU in #665
• Allow use of zonal clusters by resolving actual cluster region by @scaliby in #682
• feat: Add --sub-slicing flag to cluster create by @jamOne- in #689
• feat: Add sub_slicing_support to SystemCharacteristics by @jamOne- in #695
• feat: Bump up the default Kueue version to v0.14.1 by @jamOne- in #691
• Subslicing workload annotations by @scaliby in #712
• Release 0.14 by @FIoannides in #721

Bug fixes

• Update supported platforms by @scaliby in #663
• Pass autoprovisioning_args to Pathways workload yaml by @wstcliyu in #664
• Use map NodeSelector instead of string for Pathways workload by @wstcliyu in #669
• fix: Don't check flex in is_TAS_possible by @jamOne- in #675
• Update PathwaysJob version to v0.1.3 by @wstcliyu in #683
• Fix plain XPK command execution by @scaliby in #684
• No kueue admission checks for FSNQ by @scaliby in #686

New Contributors

• @wstcliyu made their first contribution in #664
• @RexBearIU made their first contribution in #665

Full Changelog: v0.13.0...v0.14.0

Full Changelog: v0.12.0...v0.13.0

What's Changed

New Features

• Add TAS support for workloads on DWS clusters by @FIoannides in #615
• feat: Add link to GCP Workloads list under xpk workload list by @jamOne- in #621
• feat: Make workloads link navigate to aiml jobsets list by @jamOne- in #624
• Increase kubectl wait times by @scaliby in #625
• Add pathways + nap support by @scaliby in #623

Bug fixes

• Fix error when updating a cluster that already has JobSet by @lukebaumann in #622
• Don't set --enable-queued-provisioning for flex multislice by @SikaGrr in #631

New Contributors

• @SikaGrr made their first contribution in #607
• @FIoannides made their first contribution in #615
• @jamOne- made their first contribution in #621

Full Changelog: v0.11.0...v0.12.0

What's Changed

New Features

• "Select TPU by topology (#525)" + Fix errors by @sharabiani in #563
• feat: Added an update to CoreDNS by @DannyLiCom in #530
• feat: Added an update to CoreDNS by @DannyLiCom in #501
• Add tpu7x support by @scaliby in #586
• Release v0.11.0 by @scaliby in #613

Bug fixes

• Fix max-nodes when creating flex queued nodepool of tpus by @pawloch00 in #541
• Fix kueue version in yaml string and loosen dependecy on cloud-storage by @pawloch00 in #546
• Remove RBAC container by @pawloch00 in #547
• Fix kjob.py pyink by @pawloch00 in #552
• Update Kueue to create Visibility folder by @SujeethJinesh in #556
• Update CPU limits to 750m by @SujeethJinesh in #558
• Update CPU limit for medium to large scale clusters by @SujeethJinesh in #571
• Fix cluster create when --enable-autoprovisioning is supplied by @scaliby in #589
• Fix provisioning 1x1 and 1x1x1 topologies by @scaliby in #595
• Reorder custom_nodepool_arguments for node-pool create command by @scaliby in #596
• NAP memory and cpu limit increased by @sharabiani in #597
• fix: only install JQ when not installed by @samos123 in #601
• Fix custom nodepool arguments append by @scaliby in #602
• Fix nodepool creation by @scaliby in #606

New Contributors

• @DannyLiCom made their first contribution in #530
• @scaliby made their first contribution in #581
• @samos123 made their first contribution in #601

Full Changelog: v0.10.1...v0.11.0

Bugs:

• Fixed kueue installation for tpu clusters #546

Full Changelog: v0.10.0...v0.10.1

Highlights

DWS Flex support for GPUs and TPUs
Managed Lustre storage attach support

What's Changed

New Features

• Update PathwaysJob version to v0.1.2 by @RoshaniN in #507
• Update Cluster Toolkit Version by @pawloch00 in #503
• Managed Lustre storage attach support implemented by @sharabiani in #534
• Implement DWS for GPUs and TPUs by @pawloch00 in #467

Bug fixes

• Fix issue in control_plane_endpoints_config.dns_endpoint_config.allow… by @SujeethJinesh in #499
• Fix broken A3 High workloads by @gcie in #494
• Bring back shared_memory volume for A3 Mega and A3 High by @gcie in #512
• Provided the required permissions for JAX to list the pods by @sharabiani in #509
• fix the incorrect number of chips per VM for v5litepod-8 by @gcie in #513
• Update Kueue and Jobset controller default limit value by @ycchenzheng in #502
• Fix cluster creation from reservation by @pawloch00 in #522

New Contributors

• @ycchenzheng made their first contribution in #502

Full Changelog: v0.9.0...v0.10.0

Highlights

GPUDirect-TCPX support for H100 accelerator (A3-High VMs)
A command to adapt a cluster to XPK expected config (xpk cluster adapt)
DWS Calendar Mode Reservations

What's Changed

New Features

• GPUDirect-TCPX support for H100 by @gcie in #459
• Add Multi-tier checkpointing support in XPK Cluster Creation by @abhinavclemson in #465
• Add Jobset controller patching for MTC cluster by @abhinavclemson in #475
• xpk cluster adapt by @gcie in #466

Bug fixes

• Merge main to develop by @gcie in #458
• Update pathways.py with worker component type. by @RoshaniN in #456
• Fix error when xpk storage attach --type=gcpfilestore without --mount-options by @gcie in #463
• Update README.md - text edit in Advanced usage section by @kzmyslona in #473
• Update PathwaysJob Version to v0.1.1 To Fix RM OOM by @SujeethJinesh in #477
• Placement Policy removed from A3-Mega blueprints with --spot by @sharabiani in #478
• Enable DNS Access to Prevent Connection Timeout Errors by @SujeethJinesh in #483
• Fix DWS Calendar Mode Reservations for A3 Mega by @gcie in #484

New Contributors

• @kzmyslona made their first contribution in #468

Full Changelog: v0.8.0...v0.9.0

Highlights

• Support for provisioning A4 GKE clusters
• PathwaysJob integration
• Storage support (Parallelstore and Hyperdisk)

What's Changed

New Features

• Add the option to use Multi-tier checkpointing in workloads by @abhinavclemson in #447
• Integrate PathwaysJob into XPK. by @RoshaniN in #448
• Implement Parallelstore and Hyperdisk storages attach by @sharabiani in #436
• A4 support for prod by @gcie in #412
• Add --mount-options parameter to xpk storage attach/create by @gcie in #450

Bug fixes

• Update JOBSET_VERSION from 0.7.2 to 0.8.0 by @SujeethJinesh in #425
• fix yaml alignment when remote-python-sidecar-image is passed by @sadikneipp in #426
• Bring back manual manifest specification for attaching storage by @gcie in #427
• Fix XPK version in Pypi release by @sharabiani in #428
• Remove sudo requirement from make by @sharabiani in #435
• fix: workloads not scheduling on A3 Ultra clusters by @gcie in #441
• Disable creating additional networks for L4 and A2 clusters by @gcie in #444
• Fix xpk workload create for L4 and A100 by @gcie in #452

Full Changelog: v0.7.2...v0.8.0

What's Changed

Bug fixes

• Remove sudo requirement from make by @sharabiani in #435

Full Changelog: v0.7.1...v0.7.2

What's Changed

Bug fixes

• fix yaml alignment when remote-python-sidecar-image is passed by @sadikneipp in #426
• Bring back manual manifest specification for attaching storage by @gcie in #427
• Fix XPK version in Pypi release by @sharabiani in #428