Releases · AI-Hypercomputer/xpk

09 Feb 11:17

github-actions

v1.3.0

c32f9b0

v1.3.0 Latest

Latest

What's Changed

Improvements

Record used command flags in telemetry by @scaliby in #1019
CommandsTester autopatching setup by @jamOne- in #1025
refactor: Introduce ReservationLink dataclasses and update function signatures by @jamOne- in #1029
make verify: remove installation by @jamOne- in #1033
Fix pylint by @jamOne- in #1035

Bug fixes

Fix url encoding by @scaliby in #1027

Full Changelog: v1.2.0...v1.3.0

Contributors

jamOne- and scaliby

Assets 2

02 Feb 11:26

github-actions

v1.2.0

8d4fc65

v1.2.0

What's Changed

Improvements

Replicas should be an int by @scaliby in #1013
Add PIP_OPTS by @scaliby in #1012

Bug fixes

Fix Super-slicing nume aware conflict by @jamOne- in #1001
Fix Super-slicing on 1 cube by @jamOne- in #1002
Upgrade any existing nodepools after Lustre driver installation by @SikaGrr in #1005
Set big controller resources for Super-slicing by @jamOne- in #1007
Fix workload resources in pathways and super-slicing with v7x by @jamOne- in #1009
Limit the coreDNS replica count to the desired number of default pool… by @SikaGrr in #1010

Full Changelog: v1.1.0...v1.2.0

Contributors

jamOne-, SikaGrr, and scaliby

Assets 2

23 Jan 12:04

github-actions

v1.1.2

bf04492

v1.1.2

Full Changelog: v1.1.1...v1.1.2

Assets 2

21 Jan 09:35

github-actions

v1.1.1

db607a2

v1.1.1

Full Changelog: v1.1.0...v1.1.1

Assets 2

20 Jan 14:20

github-actions

v1.1.0

80f3287

v1.1.0

What's Changed

New Features

Use gcloud beta for resource policies by @jamOne- in #962
Add multi-container support by @zxhe-sean in #880
feat: Introduce super-slicing inspection to the xpk inspector by @jamOne- in #969
Remove 1:1 workload-nodepool annotation for sub/super-slicing by @jamOne- in #970
Golden recipes script by @scaliby in #971
feat: Support multiple reservations (super-slicing sub-block targeting) by @jamOne- in #980
SUPER_SLICING_ENABLED=True by default by @jamOne- in #983
Bump Kueue to 0.15.2 and Superslicing: Remove cube state nodeLabel by @jamOne- in #990

Improvements

Support passing project number as an argument by @Aixile in #934
re-add single-host golden by @jamOne- in #944
Remove obsolete warning about single-host single-slice from documentation. by @SikaGrr in #945
remove obsolete note by @SikaGrr in #947
Remove unused makefile variables by @jamOne- in #950
Fix GPU provisioning by @FIoannides in #940
GPU e2e test by @FIoannides in #948
Docs: add "Dependencies" and "Get involved" sections by @jamOne- in #951
Fix spellcheck errors by @scaliby in #952
fix DNS credentials retry by @FIoannides in #953
Remove unused user_input.py (code was moved to console.py) by @jamOne- in #954
Add unit test for get credentials by @FIoannides in #955
Add CoreDNS to dependencies and fix the return code by @jamOne- in #958
Telemetry log is_tester by @scaliby in #957
Add note that workload create works only on XPK clusters by @scaliby in #959
Stamp telemetry trash execution by @scaliby in #963
Silent credentials check by @scaliby in #964
Add a shared memory mount for rdma_decorator by @kenmcheng in #965
Remove legacy integration tests by @scaliby in #973
Add copyright to recipes by @scaliby in #976
Allow for run only blocks in recipes by @scaliby in #979
Remove dws flex cluster ingration test as it is constantly failing by @scaliby in #972
github: Never mark issues as stale by @jamOne- in #993
Add RayCluster golden by @jamOne- in #996
Migrate to recipes by @scaliby in #994
Remove goldens script by @scaliby in #998

Bug fixes

Fix cluster adapt by @scaliby in #960
Fix RayCluster parser after enabling the SuperSlicing feature flag by @jamOne- in #988
kueue_manager: Use configure_super_slicing instead of checking system capabilities by @jamOne- in #989
Recipes exit status code by @scaliby in #995
Remove Super-slicing block validation by @jamOne- in #997
Github: remove "-nodepools" suffix from e2e cluster names by @jamOne- in #999
Fix kueuectl calls in xpk info by @jamOne- in #1000