trainer
https://github.com/kubeflow/trainer
Go
Distributed AI Model Training and LLM Fine-Tuning on Kubernetes
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Go not yet supported1 Subscribers
Add a CodeTriage badge to trainer
Help out
- Issues
- bug: Flux plugin passes per-node GPU count to -g flag instead of per-task count
- bug: RuntimePatch manager validation loses list index in error field path
- TrainJob accepts env names and resource requests that its Pods cannot use, and then never starts
- TrainJob controller ignores groupKindConcurrency and always runs one worker
- A TrainJob with a negative resourcesPerNode request is admitted and crash-loops the Volcano scheduler
- fix(controller): disable HTTP/2 for TLS configs that cannot serve it
- OptimizationJob with DRA fields in trainJobTemplate is admitted while TrainJobDynamicResourceAllocation is disabled, then never creates a trial
- chore(docs): add add-training-runtime skill
- fix(initializer): use watch API to wait for the cache LeaderWorkerSet
- fix(api): show accurate TrainJob State after resume and add Suspended printer column
- Docs
- Go not yet supported