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
- fix(controller): clamp activeDeadlineSeconds to avoid duration overflow
- fix(controller): clamp ActiveDeadlineSeconds before converting to time.Duration
- chore(runtimes): add FSDP and DeepSpeed TrainingRuntime guide
- fix: cover TrainJob lookup failure in cache initializer
- fix: cover TrainJob lookup failure
- feat(initializers): fall back to bin/pt/pth or gguf weights when a HF model repo has no safetensors
- fix(initializer): cache initializer waits for the LeaderWorkerSet with an unbounded poll loop
- fix(controller): validate and reject conflicting runtime volume and volumeMount in TrainJobStatus
- Coscheduling and Volcano plugins drop the missing PodGroup CRD error
- bug: Flux shared-memory volume mounted read-only, preventing writes to /dev/shm
- Docs
- Go not yet supported