Pipelines: Unable to visualize table in OutputViewer

Created on 6 Dec 2018  路  14Comments  路  Source: kubeflow/pipelines

I am trying to visualize some artifacts in the OutputViewer with no success.
I have the following metadata.json file in the root of the container's file system:

{
    "version": 1,
    "outputs": [{
        "type": "table",
        "source": "/linear_model_metrics.csv",
        "header": ["Epoch", "Loss_te", "Acc", "Loss_tr"],
        "format": "csv"
    }]
}

The file linear_model_metrics.csv is in the file system root as well. _Comma separated_ with first line as header. In the Pipelines UI I have a blank page in the Artifacts tab, no errors thrown.

arecrviewer help wanted kinfeature needs investigation prioritp1

Most helpful comment

+1 on supporting local mount. This is absolutely necessary for on-prem deployment. I'm loading the data to an nfs volume in preprocessing, training steps ... would be great to read it to tensor-board directly from there.

All 14 comments

Currently, the Pipelines UI reads the artifact data from the storage service directly, and only GCS and minio are supported at the moment, adding support for more services is welcome of course. If your data is on GCS, the source field should be a gs://... path.

We are currently exploring better ways to mount/read this data such that it's easier to keep the data in the cluster, or in other types of storage.

/cc @Ark-kun.

Thank you for the response. For sure it would be helpful to be able to load files directly from the container's local file system. Any info regarding the timeline or priorities on these features?

@StefanoFioravanzo Just wondering did you mount the volume through the container Op, and writes linear_model_metrics.csv to it directly?

@IronPan I mounted the volume like you suggested in #477 . And yes I wrote linear_model_metrics.csv directly form the container.

Is this still the case?
Does the source field have to be either in GCS or minio?

+1 on supporting local mount. This is absolutely necessary for on-prem deployment. I'm loading the data to an nfs volume in preprocessing, training steps ... would be great to read it to tensor-board directly from there.

+1 for supporting local storage mounted from PersistentVolume.

+1 for supporting local storage mounted from PersistentVolume or Volume

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.

Any update on this? +1 for support for local sources

/reopen

@Bobgy: Reopened this issue.

In response to this:

/reopen

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

https://github.com/kubeflow/pipelines/pull/4236 supports this feature request.

Was this page helpful?
0 / 5 - 0 ratings