[
https://issues.apache.org/jira/browse/SUBMARINE-66?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Szilard Nemeth updated SUBMARINE-66:
------------------------------------
Attachment: SUBMARINE-66.002.patch
> Improve TF config env JSON generator + tests
> --------------------------------------------
>
> Key: SUBMARINE-66
> URL: https://issues.apache.org/jira/browse/SUBMARINE-66
> Project: Hadoop Submarine
> Issue Type: Improvement
> Reporter: Szilard Nemeth
> Assignee: Szilard Nemeth
> Priority: Major
> Attachments: SUBMARINE-66.001.patch, SUBMARINE-66.002.patch
>
>
> org.apache.hadoop.yarn.submarine.runtimes.yarnservice.tensorflow.TensorFlowCommons#getTFConfigEnv
> generates a JSON of the TF_CONFIG environment variable.
> This code could be improved to use a JSON serializer instead of hand-crafting
> JSON data.
> The test class of this class also could be improved:
> org.apache.hadoop.yarn.submarine.runtimes.yarnservice.TestTFConfigGenerator
> In this class, there are some quite unreadable JSON strings, this could be
> refactored to be read out from a file instead.
> TestTFConfigGenerator has only one testcase and it's ugly, as we use very
> long strings to verify the JSON data produced matches our expected JSON.
> We should use JSON files instead of manually constructing Strings in tests,
> especially if they are very long.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
