[ https://issues.apache.org/jira/browse/MADLIB-1417?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Frank McQuillan closed MADLIB-1417. ----------------------------------- Resolution: Fixed This is not an actual issue. The test failures were probably caused by the search_path of dev check. `public` is not included in the `search_path` which means the predict, assert etc cannot find the output table. Now use pg_temp instead of public schema for dev-check https://github.com/apache/madlib/commit/404e62381248f81f84ec6032a86e67c9b215d620 > DL: Support schema qualified output tables for fit and fit_multiple > ------------------------------------------------------------------- > > Key: MADLIB-1417 > URL: https://issues.apache.org/jira/browse/MADLIB-1417 > Project: Apache MADlib > Issue Type: Bug > Components: Deep Learning > Reporter: Nikhil Kak > Assignee: Orhan Kislal > Priority: Minor > Fix For: v1.17 > > > dev-check fails if the output table is schema qualified. > I modified the madlib_keras_transfer_learning dev-check test by schema > qualifying the output table argument in fit multiple > {code:java} > SELECT madlib_keras_fit_multiple_model( > 'iris_data_packed', > 'public.iris_multiple_model', > 'mst_table', > 3, > FALSE, NULL, 1 > ); {code} > and > {code:java} > SELECT madlib_keras_fit_multiple_model( > 'iris_data_packed', > 'public.iris_multiple_model', > 'mst_table', > 3, > FALSE, > NULL, 1, > TRUE -- warm_start > ); {code} > > Both of these cases fail because of different reasons. > It is also likely that madlib_keras_fit is also broken when a schema > qualified output table is passed in. We should add tests for all the dl > functions that create an output table and fix all the failures. > -- This message was sent by Atlassian Jira (v8.3.4#803005)