mirror of https://github.com/kubeflow/trainer.git
* Remove TrainJobCreated condition Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Update KEP-2170 proposal Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Remove Created condition from SDK Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Default TrainJob status to Created unconditionally Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Set Failed condition on TrainJob runtime creation errors Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Emit a warning event upon TrainJob resources reconcile error Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Update TrainJob resources creation failed event Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Truncate event message to the maximum length limit Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Update state diagram in KEP-2170 Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> * Append ellipsis to event message if it's truncated Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> --------- Signed-off-by: Antonin Stefanutti <antonin@stefanutti.fr> |
||
---|---|---|
.. | ||
2003-train-api | ||
2145-jax-integration | ||
2170-kubeflow-trainer-v2 | ||
2401-llm-trainer-v2 | ||
README.md |
README.md
Proposals
Kubeflow uses the KEP process to document large scale changes to the project.
Details on the process (including the KEP template, recommendations, etc.) can be found at kubeflow/community/proposals