Adds a Tensorboard component for local environments #792
Conversation
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: If they are not already assigned, you can assign the PR to them by writing The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
1 similar comment
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: If they are not already assigned, you can assign the PR to them by writing The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
requested review from
gaoning777,
IronPan,
hongye-sun and
vicaire
|
/assign @IronPan |
There was a problem hiding this comment.
Reviewed 4 of 4 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @gaoning777, @hongye-sun, @IronPan, @swiftdiaries, and @vicaire)
|
@neuromage added a CRD to start and manage instances of Tensorboard. The advantage of this CRD is that it manages the tensorboard instances and keeps track of them. https://github.com/kubeflow/pipelines/tree/master/backend/src/crd/samples/viewer Could we add the functionality of this PR to this CRD? It should then be easy to create an instance of Tensorboard by using kubectl. Doing this will also put us on the right track to integrate this functionality with the UI. /cc @neuromage |
|
@animeshsingh Thank you for the review :) |
|
@vicaire @neuromage Thank you for the reply. The ViewerCRD looks pretty neat. How does Pipelines run / orchestrate the ViewerCRD and how can it be attached to a training ContainerOp / TfJob op for example? |
|
/cc @neuromage @swiftdiaries, the CRD does not yet support mounting a volume. Would you be interested in adding support for this? The way it works today is like this:
WDYT? |
|
Yeah, the viewer CRD supports volumes. I tested this but should have documented and added a sample for how to do it. Thanks @swiftdiaries for looking into it and adding the sample. The next step is to replace the Tensorboard launcher in the UI to use the viewer CRD instead. If you'd like to look into that @swiftdiaries , that would be great. /cc @yebrahim |
|
Thank you :) I’ll look into it and send a PR shortly
…Sent from my iPhone
On Feb 13, 2019, at 23:23, Ajay Gopinathan ***@***.***> wrote:
Yeah, the viewer CRD supports volumes. I tested this but should have documented and added a sample for how to do it. Thanks @swiftdiaries for looking into it and adding the sample.
The next step is to replace the Tensorboard launcher in the UI to use the viewer CRD instead. If you'd like to look into that @swiftdiaries , that would be great.
/cc @yebrahim
/cc @rileyjbauer
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.
|
|
SGTM. Thanks. Should we close this PR? |
Right now, the Output Artifacts Viewer (Pipelines UI) only accepts GCS and Minio as the storage services Source Code.
However, Tensorboard cannot process links starting with
minio://and as of now there's no way to read things off of the local filesystem through volumes and/or volume mounts.This is a component that we have been using for our onprem demos, which we are working on open-sourcing.
This change is