Deeplearning4j: Remote UI and spark conflict issues: Need better documentation

Created on 27 Jul 2017  路  2Comments  路  Source: eclipse/deeplearning4j

Spark conflicts with our UI forcing the need for a remote UI. This is covered rather well (I think) in the docs under distributed training, however it seems they could be re-organized or clarified in some way. Opening this issue for a general revue and clarification of the UI documentation to verify it is well covered in a central easy to find location.

I will close https://github.com/deeplearning4j/deeplearning4j/issues/2617

And refer all discussion to this issue.

Documentation

Most helpful comment

I tried AnimalsClassification from dl4j-examples and had the same error as is described in #2617.
I dont need Spark. So I removed this dependency:

      <groupId>org.datavec</groupId>"
      <artifactId>datavec-spark_${scala.binary.version}</artifactId>

from my pom.xml and example is working.

Is any possibility not remove Spark dependency and "say" (set): Now I dont want to use Spark?

Thank you.

All 2 comments

Unfortunately, the UIExample crashes for the same reason. How's one supposed to fix this?

Moreover, why does the doc explicitly mentions that the remote UI functionality "will require the deeplearning4j-ui_2.10 or deeplearning4j-ui_2.11 dependency"? Doesn't step one of the tutorial says it's a prerequisite anyway?

I tried AnimalsClassification from dl4j-examples and had the same error as is described in #2617.
I dont need Spark. So I removed this dependency:

      <groupId>org.datavec</groupId>"
      <artifactId>datavec-spark_${scala.binary.version}</artifactId>

from my pom.xml and example is working.

Is any possibility not remove Spark dependency and "say" (set): Now I dont want to use Spark?

Thank you.

Was this page helpful?
0 / 5 - 0 ratings