 Andrey VelichkevichandGitHub
|
fbacd9f577
|
fix(examples): Verify TrainJob Completion (#3331)
* fix(examples): Verify TrainJob Completion
Signed-off-by: Andrey Velichkevich <[email protected]>
* Rename client to TrainerClient()
Signed-off-by: Andrey Velichkevich <[email protected]>
---------
Signed-off-by: Andrey Velichkevich <[email protected]>
|
2026-03-16 14:40:31 +00:00 |
|
|
Akash JaiswalandGitHub
|
9a08b4e47f
|
chore: migrate to a10.2 gpu for gpu e2e (#3220)
Signed-off-by: Akash Jaiswal <[email protected]>
|
2026-02-18 09:16:29 +00:00 |
|
|
Khushi AgrawalandGitHub
|
60db275023
|
feat(runtimes): add support for ClusterTrainingRuntimes in Helm chart (#3124)
* feat(runtimes): add support for ClusterTrainingRuntimes in Helm chart
Signed-off-by: khushiiagrawal <[email protected]>
* fix: remove initializerImage from user-configurable values
Signed-off-by: khushiiagrawal <[email protected]>
* chore: regenerate README with helm-docs
Signed-off-by: khushiiagrawal <[email protected]>
* fix: address Copilot review suggestions
Signed-off-by: khushiiagrawal <[email protected]>
* feat: Introduce helper to centralize image tag resolution
Signed-off-by: khushiiagrawal <[email protected]>
* refactor: nest cache image configuration and update copyright year.
Signed-off-by: khushiiagrawal <[email protected]>
* chore: run make generate to sync
Signed-off-by: khushiiagrawal <[email protected]>
* feat: add TorchTune distributed runtime, image helper usage, and default runtimes enabled flag
Signed-off-by: khushiiagrawal <[email protected]>
* feat: enable runtime via a new default flag and add a comment
Signed-off-by: khushiiagrawal <[email protected]>
* refactor: Torchtune runtimes to use model specific configurations, add default enabled option
Signed-off-by: khushiiagrawal <[email protected]>
* fix: update README and fix trailing whitespace for CI
Signed-off-by: khushiiagrawal <[email protected]>
* refactor: relocate runtime template and update its configuration
Signed-off-by: khushiiagrawal <[email protected]>
* feat: add JAX distributed training support and update runtime configurations
Signed-off-by: khushiiagrawal <[email protected]>
* fix(docs): remove JAX runtime from default runtimes in README
Signed-off-by: khushiiagrawal <[email protected]>
* fix(docs): update default runtimes in README and values.yaml to include JAX
Signed-off-by: khushiiagrawal <[email protected]>
* fix(workflows): increase Papermill timeout for e2e tests in GPU cluster
Signed-off-by: khushiiagrawal <[email protected]>
* fix(notebooks): add missing newline at end of qwen2.5-1.5B-with-alpaca.ipynb
Signed-off-by: khushiiagrawal <[email protected]>
---------
Signed-off-by: khushiiagrawal <[email protected]>
Signed-off-by: Khushi Agrawal <[email protected]>
|
2026-02-15 16:15:27 +00:00 |
|
|
Andrey VelichkevichandGitHub
|
2612533a5d
|
feat(runtimes): Use JobSet VolumeClaimPolicies APIs for LLM Runtimes (#3150)
* feat(runtimes): Use JobSet VolumeClaimPolicies APIs for LLM Runtimes
Signed-off-by: Andrey Velichkevich <[email protected]>
* Remove cell
Signed-off-by: Andrey Velichkevich <[email protected]>
---------
Signed-off-by: Andrey Velichkevich <[email protected]>
|
2026-02-09 14:57:13 +00:00 |
|
|
Akash JaiswalandGitHub
|
dc4267f64e
|
feat: replaced vm runner with test gpu arc from cncf (#3067)
* test: gpu based ARC from cncf instead of stanalone vm
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: removed label for gpu e2e test
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: remove delete cluster action
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: rm seperate script, rm path and nvidia smi command for testing
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: using nvkind with sudo
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: move nvkind from noexec to local/bin
Signed-off-by: Akash Jaiswal <[email protected]>
* tmp: install nvkind as from arc is not wroking
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: run the commands as sudo (check https://github.com/NVIDIA/nvkind/issues/48)
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: nvkind as sudo
Signed-off-by: Akash Jaiswal <[email protected]>
* test: ignore the patch error
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: downgrade the version for nvidia ctk
Signed-off-by: Akash Jaiswal <[email protected]>
* add: service restart
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: patch version for nctk
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: command
Signed-off-by: Akash Jaiswal <[email protected]>
* refactor: split into different script
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: transfer to legacy mode
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: legacy mode for nctk
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: downgrade nctk
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: non root kubectl
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: helm dirs
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: refactored the code
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: rm seperate script, migrated to prod gpu arc, and other misc
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: rm sudo for kind
Signed-off-by: Akash Jaiswal <[email protected]>
* add: wait fo qwen to complete
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: fix qwen nb and fix gpu operator to use host driver
Signed-off-by: Akash Jaiswal <[email protected]>
* test: use default value for ntk toolkit
Signed-off-by: Akash Jaiswal <[email protected]>
* hotfix: patch CRDs to run on GPU nodes (Check #3067)
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: the patching to mdoel and initializers
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: single node for gpu e2e test
Signed-off-by: Akash Jaiswal <[email protected]>
* revert: jax eg
Signed-off-by: Akash Jaiswal <[email protected]>
---------
Signed-off-by: Akash Jaiswal <[email protected]>
|
2026-02-08 21:07:13 +00:00 |
|
|
Anya KramarandGitHub
|
ea8b4b596a
|
chore: Install released version of Kubeflow SDK (#2857)
Signed-off-by: kramaranya <[email protected]>
|
2025-09-29 09:23:31 +00:00 |
|
|
Akash JaiswalandGitHub
|
cbb653e620
|
feat: qwen 2.5 1.5b runtime, example and fix gpu e2e test (#2835)
* add: qwen 3 1.7b CTR
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: pvc and ctr
Signed-off-by: Akash Jaiswal <[email protected]>
* feat: qwen 2, 1.5b model as example
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: allow pattern and config update
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: test and typo
Signed-off-by: Akash Jaiswal <[email protected]>
* update: qwen3 model
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: add qwen 2.5, fix gpu e2e test and use local images
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: whitespace
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: review suggestions
Signed-off-by: Akash Jaiswal <[email protected]>
* fix
Signed-off-by: Akash Jaiswal <[email protected]>
---------
Signed-off-by: Akash Jaiswal <[email protected]>
|
2025-09-20 21:02:59 +00:00 |
|
|
Akash JaiswalandGitHub
|
82ad2deb74
|
feat: support for managing gpu enabled self runner infra (#2762)
* feat: support for creating and managing gpu cluster
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: makefile bug
Signed-off-by: Akash Jaiswal <[email protected]>
* add: ci action to ask maintainers to add label to when changes are detected
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: fixed issues and cleanup
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: run check on change in pr
Signed-off-by: Akash Jaiswal <[email protected]>
* feat: added seperate workflow for gpu runner
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: deepspeed typo
Signed-off-by: Akash Jaiswal <[email protected]>
* hotfix: add gpu label on PR without merging
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: merged into single action
Signed-off-by: Akash Jaiswal <[email protected]>
* fixL run runner as soon as label is added
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: use gpu runner when label exist
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: revert changes and fix script permission
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: create gpu supported gpu
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: nvidia issue
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: gpu cluster and torchtune model
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: notebookpath and delete cluster
Signed-off-by: Akash Jaiswal <[email protected]>
* tmp fix: notebook to use k8s client
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: use akash sdk and fix notenook size
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: notebook error
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: delete cluster before creating one and notebook
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: kube config
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: makefile add comment
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: nvidia runtime
Signed-off-by: Akash Jaiswal <[email protected]>
* hotfix: disable e2e go
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: delete cluster
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: delete cluster
Signed-off-by: Akash Jaiswal <[email protected]>
* hotfix: temporarly use my personal token
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: refactored code
Signed-off-by: Akash Jaiswal <[email protected]>
* hotfix: take hf token from env of self runner vm
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: to run notebook directly
Signed-off-by: Akash Jaiswal <[email protected]>
* refactor: torchtune job
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: ci action
Signed-off-by: Akash Jaiswal <[email protected]>
* fix: pre commit hook
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: rename ci action
Signed-off-by: Akash Jaiswal <[email protected]>
* rem: delete cluster command from makefile
Signed-off-by: Akash Jaiswal <[email protected]>
* chore: rem some steps, fixed wait timing and notebook logs according to kubeflow/sdk#83
Signed-off-by: Akash Jaiswal <[email protected]>
* update: upgrade k8s to 1.34.0
Signed-off-by: Akash Jaiswal <[email protected]>
---------
Signed-off-by: Akash Jaiswal <[email protected]>
|
2025-09-04 15:33:12 +00:00 |
|
|
Anya KramarandGitHub
|
c83e6db3c6
|
fix: update kubeflow sdk reference (#2780)
Signed-off-by: kramaranya <[email protected]>
|
2025-08-07 15:13:56 +00:00 |
|
|
Andrey VelichkevichandGitHub
|
dba2af939a
|
fix(examples): Update the argument for Runtime framework (#2766)
* fix(examples): Update the runtime framework parameter
Signed-off-by: Andrey Velichkevich <[email protected]>
* Update TorchTune example
Signed-off-by: Andrey Velichkevich <[email protected]>
* Use wait_for_job_status API
Signed-off-by: Andrey Velichkevich <[email protected]>
* Fix Notebook
Signed-off-by: Andrey Velichkevich <[email protected]>
* Fix client
Signed-off-by: Andrey Velichkevich <[email protected]>
* Fix job name
Signed-off-by: Andrey Velichkevich <[email protected]>
---------
Signed-off-by: Andrey Velichkevich <[email protected]>
|
2025-08-04 14:29:53 +00:00 |
|
|
Shao WangandGitHub
|
d27e0a0498
|
feat(example): Add alpaca-trianjob-yaml.ipynb. (#2670)
* feat(example): Add alpaca-trianjob-yaml.ipynb.
Signed-off-by: Electronic-Waste <[email protected]>
* fix(example): Update the overview of the torchtune llama3_2 example.
Signed-off-by: Electronic-Waste <[email protected]>
* fix(example): Update the pvc description.
Signed-off-by: Electronic-Waste <[email protected]>
* chore(example): Add the get the fine-tuned model section.
Signed-off-by: Electronic-Waste <[email protected]>
* fix(example): Fix some errors.
Signed-off-by: Electronic-Waste <[email protected]>
* fix(example): fix some errors.
Signed-off-by: Electronic-Waste <[email protected]>
* fix(manifests): Fix debug tag.
Signed-off-by: Electronic-Waste <[email protected]>
* fix(example): Change PVC creation method to Python SDK.
Signed-off-by: Electronic-Waste <[email protected]>
* fix(example): Remove config load.
Signed-off-by: Electronic-Waste <[email protected]>
---------
Signed-off-by: Electronic-Waste <[email protected]>
|
2025-07-02 17:16:18 +00:00 |
|