Skip to content

Commit

Permalink
katib metrics-collector: mention supported writers
Browse files Browse the repository at this point in the history
See kubeflow/katib#2467

Signed-off-by: Gary Miguel <[email protected]>
  • Loading branch information
garymm authored Feb 13, 2025
1 parent a8d15e9 commit c43c19e
Showing 1 changed file with 3 additions and 0 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,9 @@ To define the pull-based metrics collector for your Experiment:

- `TensorFlowEvent`: Katib collects the metrics from a directory path
containing a [tf.Event](https://www.tensorflow.org/api_docs/python/tf/compat/v1/Event).
These are typically written by [tensorflow.summary](https://www.tensorflow.org/api_docs/python/tf/summary).
As of Katib 0.18, [torch.utils.tensorboard](https://pytorch.org/docs/stable/tensorboard.html) or
[tensorboardX](https://tensorboardx.readthedocs.io/en/latest/index.html) may also be used to write.
You should specify the path in the `.source.fileSystemPath.path` field. Check the
[TFJob example](https://github.com/kubeflow/katib/blob/ea46a7f2b73b2d316b6b7619f99eb440ede1909b/examples/v1beta1/kubeflow-training-operator/tfjob-mnist-with-summaries.yaml#L17-L23).
The default directory path is `/var/log/katib/tfevent/`.
Expand Down

0 comments on commit c43c19e

Please sign in to comment.