[
https://issues.apache.org/jira/browse/SOLR-11838?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16322824#comment-16322824
]
Christine Poerschke commented on SOLR-11838:
--------------------------------------------
Hi, thanks everyone for your comments so far here, that's exactly and more than
what I'd hoped for when putting 'explore' in the ticket title :-) I'll remove
the 'in contrib/ltr' suffix to reflect the general scope.
bq. ... Pre-trained models ... retrieve them and hydrate them with
ModelSerializer.restoreMultiLayerNetwork(InputStream) ...
"Retrieve a pre-trained model, hydrate it and use it (for contrib/ltr)." - that
would nicely capture an aspect I'm personally particularly interested in.
Thanks for the pointer to
{{ModelSerializer.restoreMultiLayerNetwork(InputStream)}}!
At present the LTR models are stored in a ManagedModelStore/ManagedResource
i.e. in ZooKeeper; blob store use has been explored but as yet it's not yet
used. With SOLR-11250 there is now also provision for the model definition to
be essentially outside the model store with the store stored content just
naming the .json file that has the actual model:
https://lucene.apache.org/solr/guide/7_2/learning-to-rank.html#using-large-models
Also on the subject of model storage/retrieval, on [~kaessmann]'s SOLR-9887
(about reading KeepWordFilter etc. data from a JDBC source) [~dsmiley] mentioned
bq. ... it'd be cool if these might pull from a StreamingExpression, which in
turn could be an expression that reads from a JDBC or Solr or who knows what
else someone might plug in.
and whilst word filter data and machine learning models are kind of different,
the idea of Solr using a streaming expression to pull data from <somewhere>
seems fairly generic and very appealing?
> explore supporting Deeplearning4j NeuralNetwork models in contrib/ltr
> ---------------------------------------------------------------------
>
> Key: SOLR-11838
> URL: https://issues.apache.org/jira/browse/SOLR-11838
> Project: Solr
> Issue Type: New Feature
> Components: contrib - LTR
> Reporter: Christine Poerschke
> Attachments: SOLR-11838.patch
>
>
> [~yuyano] wrote in SOLR-11597:
> bq. ... If we think to apply this to more complex neural networks in the
> future, we will need to support layers ...
> [~malcorn_redhat] wrote in SOLR-11597:
> bq. ... In my opinion, if this is a route Solr eventually wants to go, I
> think a better strategy would be to just add a dependency on
> [Deeplearning4j|https://deeplearning4j.org/] ...
> Creating this ticket for the idea to be explored further (if anyone is
> interested in exploring it), complimentary to and independent of the
> SOLR-11597 RankNet related effort.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
