...
Use the ML Job studio plugin to run training and inference jobs. To run the jobs on data which are indexed in Squirro we suggest to use the SquirroQueryLoader. Further to store the classifications and predictions we suggest for sentence-level jobs to use the SquirroEntityFilter step and for document level jobs to use the SquirroItemSaver step.
If the a trained model is uploaded with the custom classifier step, we can also make it available in the enrich pipeline to directly classify documents which get loaded . Currently there is not straight way to create a Publish Model directly from a custom ML Workflow via the SquirroClient. So we would need to access the
machinelearning
DB and update theworkflow_id
withid
of the custom ml workflow in a existing row in thepublish
table. Check the workflow section of the ML Template to see which steps are required in the ML workflow(How-to Publish ML Models Using the Squirro Client ).The last option is to make a custom step available within a ML Template in the AI Studio. Currently there is no other option then transform the custom step to a actual step of the lib/nlp and then creating a new ML template by following the documentation (https://nektoon.atlassian.net/wiki/spaces/SQ/pages/1870955131).