area/examplesgood first issuehelp wanted
倉庫指標
- 星標
- (2,112 顆星)
- PR 合併指標
- (平均合併 10天 9小時) (30 天內合併 75 個 PR)
描述
It would be good to run GPU E2E tests for DeepSpeed and MLX example Notebooks to tests our MPI-based runtimes.
- https://github.com/kubeflow/trainer/blob/master/examples/mlx/language-modeling/fine-tune-llama.ipynb
- https://github.com/kubeflow/trainer/blob/master/examples/deepspeed/text-summarization/T5-Fine-Tuning.ipynb
That should help us detect MPI errors earlier, like here: https://github.com/kubeflow/trainer/pull/3748
/good-first-issue /area examples