AI-Hypercomputer xpk issues

AI-Hypercomputer / xpk

xpk (Accelerated Processing Kit, pronounced x-p-k,) is a software tool to help Cloud developers to orchestrate training jobs on accelerators such as TPUs and GPUs on GKE.

Apache License 2.0

81 stars 23 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Add logic to fail Pathways jobs on user code errors.

#152 RoshaniN closed 5 months ago
1
Add a check and update existing Pathways clusters to use CloudDNS.

#151 RoshaniN closed 5 months ago
0
Move all clusters to be RAPID clusters, and verify them using valid_v…

#150 Obliviour closed 5 months ago
0
Fixing device_type in nightly tests.

#149 RoshaniN closed 5 months ago
0
Restrict Pathways unified debugging logs to just first worker

#148 SujeethJinesh closed 5 months ago
0
Fix stacktrace sidecar container yaml

#147 SurbhiJainUSC closed 5 months ago
0
Don't Kill RM or Proxy on user job failure

#146 SujeethJinesh opened 6 months ago
0
Enable cluster and workload creation on A3+.

#145 yangyuwei closed 5 months ago
1
Add Unified Logging View for Pathways on Cloud

#144 SujeethJinesh closed 6 months ago
1
Update pathways server and proxy server image locations

#143 sadikneipp closed 6 months ago
0
Fixing misleading message on Validate Docker Image.

#142 RoshaniN closed 6 months ago
0
Pathways in headless mode.

#141 RoshaniN closed 5 months ago
0
Allow JAX coordinator to find the JobSet name.

#140 RoshaniN closed 6 months ago
0
Making exit flow similar to other XPK commands.

#139 RoshaniN closed 6 months ago
1
Update pip version to 0.5.0

#138 SurbhiJainUSC closed 7 months ago
1
Update xpk.py

#137 kyle-google closed 6 months ago
2
Remove user-managed service account and attach default compute engine service account to node pools

#136 SurbhiJainUSC closed 7 months ago
0
Dynamically determine GKE Version for Cluster and Node Pool Creation

#135 Obliviour closed 7 months ago
0
Prevent Pathways SIGTERMs from counting against backoffLimit

#134 SujeethJinesh opened 7 months ago
0
enable create workload for h150

#133 NinaCai opened 7 months ago
0
Fix incorrect indent in workload list output

#132 Obliviour closed 7 months ago
0
Disable service account feature

#131 SurbhiJainUSC closed 7 months ago
1
Set gcloud zone property for build and nightly tests

#130 SurbhiJainUSC closed 7 months ago
0
Add project flag to service account commands

#129 SurbhiJainUSC closed 7 months ago
0
Add project flag to service account commands and add random gcloud properties to integ tests

#128 SurbhiJainUSC closed 7 months ago
0
Add Support for Pathways Expected Instances & Larger Default Worker Backoff Limit

#127 SujeethJinesh closed 7 months ago
0
Import cloud-accelerator-diagnostics only when Vertex AI Tensorboard flag is set

#126 SurbhiJainUSC closed 7 months ago
0
Add configuration setting for default pool num nodes

#125 Obliviour closed 7 months ago
0
Add custom env variables to CPU workloads.

#124 RoshaniN closed 7 months ago
0
Enable formating with pyink to adhere with google3 style.

#123 Obliviour closed 7 months ago
0
Update README with Vertex AI Tensorboard information and update pip version to 0.4.0

#122 SurbhiJainUSC closed 7 months ago
1
XPK cleanup: integ tests and code cleanup

#121 Obliviour opened 7 months ago
0
Check cluster arguments and update nodepools in existing cluster when requesting different device_type

#120 SurbhiJainUSC closed 6 months ago
0
Fix None docker_name and wait-for-workload-completition poll mode

#119 Obliviour closed 7 months ago
0
Revert "Nina/unify gpu container yaml"

#118 NinaCai closed 7 months ago
2
Add timeout=0 to readme

#117 raymondzouu closed 7 months ago
0
Add wait-for-job-completion to integration test

#116 raymondzouu closed 7 months ago
0
Nina/unify gpu container yaml

#115 NinaCai closed 7 months ago
0
CPU shared clusters for Llama and Mistral model runs.

#114 RoshaniN closed 7 months ago
0
Add warning when user schedules workload on a cluster created using previous XPK version

#113 SurbhiJainUSC closed 7 months ago
0
return pid exit code if it is non-zero

#112 NinaCai closed 7 months ago
0
Delete subnets when deleting the cluster

#111 NinaCai closed 7 months ago
0
Change tensorboard_location to tensorboard_region for compatibility

#110 SurbhiJainUSC closed 7 months ago
0
Retry again with longer wait times for kueue credentials step

#109 Obliviour closed 7 months ago
0
Add --second-docker-image option

#108 tonyjohnchen closed 7 months ago
2
Add workload list wait-for-job-completion feature

#107 raymondzouu closed 7 months ago
0
Enable Autoprovisioning Support in XPK

#106 Obliviour closed 7 months ago
0
Add dynamic versioning for pip package

#105 SurbhiJainUSC closed 7 months ago
0
Support --env flag and Artifact Registry image validation

#104 jonb377 closed 7 months ago
0
Add Pathways end-to-end tests to build tests and nightly tests.

#103 RoshaniN closed 7 months ago
0

Previous Next