Tf-operator: [feature] Support multi port in TFJob

Created on 23 Feb 2021  路  8Comments  路  Source: kubeflow/tf-operator

Now we do not support multi ports in TFJob because we do not expose them in the service.

kinfeature

Most helpful comment

/PTAL @gaocegege Can i start this work?

All 8 comments

/assign @xieydd

ok, i will do it.

Hello, any update on this feature? Without multi-port support, it's harder to enable Profiler integration for multiple workers using TensorFlow 2.4+. Thanks!

@xieydd thanks for merging the feature in kubeflow/common. Does tf-operator need to update its dependency for this to work? And what's tf-operator's release schedule?

Does tf-operator need to update its dependency for this to work

Yeah.

what's tf-operator's release schedule?

We release a new version with a kubeflow release regularly.

After we merge #1276, I think we can work on this issue.

what's tf-operator's release schedule?

We release a new version with a kubeflow release regularly.

After we merge #1276, I think we can work on this issue.

/cc @gaocegege @verult l think we could proceed on this due to #1276 was merged.

/PTAL @gaocegege Can i start this work?

I think we can, wdyt

Was this page helpful?
0 / 5 - 0 ratings

Related issues

pingsutw picture pingsutw  路  3Comments

zuowang picture zuowang  路  7Comments

dpaks picture dpaks  路  5Comments

pingsutw picture pingsutw  路  7Comments

andreyvelich picture andreyvelich  路  9Comments