opendatahub-io / kubeflow-sdk / 22845728048

76%

Build Type

push

github

Committed by web-flow

Commit Message

feat: Automatic final model upload to S3 after training (#85)

* feat: add automatic final model upload on save_model() call

- Monkey-patch Trainer.save_model() to automatically upload final model to S3
- Supports all distributed strategies: DDP, FSDP (sharded/full), DeepSpeed
- Uses barrier sync to prevent race conditions in FSDP sharded mode
- Reuses existing upload queue infrastructure

* fix: restart upload worker when thread dies

* chore: update async upload funcs to be public

* test: add unit tests for final model feat.

* feat: upload final model artifacts as post training clean up code

Run Details

208 of 272 new or added lines in 2 files covered. (76.47%)

1 existing line in 1 file now uncovered.

4988 of 7262 relevant lines covered (68.69%)

1.37 hits per line