[ https://issues.apache.org/jira/browse/IGNITE-11137?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Anton Dmitriev updated IGNITE-11137: ------------------------------------ Description: Currently we have integration between machine learning and SQL implemented in IGNITE-11138, IGNITE-11071 and IGNITE-11072. This functionality allows to work with model storage in a straight forward way, user can save model without any checking so that it might be overridden; model is extracted on each predict call and it's very inefficient. The goal of this task is to: * Add existence checking to model saving functionality and meaningful exception messages; * Add model caching into predict call so that model is not required to be deserialized on each call. was: We want to wrap Model storage by IgniteModelStorage. This wrapper should: * hide all serialization/deserialization activities for models * check args and work with paths which not exist yet * cache used models from storage > [ML] IgniteModelStorage > ----------------------- > > Key: IGNITE-11137 > URL: https://issues.apache.org/jira/browse/IGNITE-11137 > Project: Ignite > Issue Type: Improvement > Components: ml > Reporter: Yury Babak > Assignee: Anton Dmitriev > Priority: Major > Fix For: 2.8 > > > Currently we have integration between machine learning and SQL implemented in > IGNITE-11138, IGNITE-11071 and IGNITE-11072. This functionality allows to > work with model storage in a straight forward way, user can save model > without any checking so that it might be overridden; model is extracted on > each predict call and it's very inefficient. The goal of this task is to: > * Add existence checking to model saving functionality and meaningful > exception messages; > * Add model caching into predict call so that model is not required to be > deserialized on each call. -- This message was sent by Atlassian JIRA (v7.6.3#76005)