Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support loading rankers from other elasticsearch plugins #202

Open
ebernhardson opened this issue Dec 12, 2018 · 2 comments
Open

Support loading rankers from other elasticsearch plugins #202

ebernhardson opened this issue Dec 12, 2018 · 2 comments

Comments

@ebernhardson
Copy link
Collaborator

Elasticsearch contains a class, ExtensiblePlugin, that allows a plugin to define an spi and then load classes from other plugins. There is example code for doing this in the painless extension. If we want to support JNI rankers, such as tensorflow, in the future having them as separate plugins would be desirable.

@karsures
Copy link

karsures commented Sep 20, 2019

Hi @ebernhardson @epugh @sstults @purbon @jettro @o19s-admin @softwaredoug ,

Are there any updates regarding this? Is there an alternate way to convert TF models into ES-LTR ranker compatible format?

@ebernhardson
Copy link
Collaborator Author

I previously hacked up the plugin to test run models through tensorflow[1]. While interesting the complications of JNI along with performance concerns (we could only batch model evaluation per segment, and in our prod use case 1 result per segment is reasonably common) basically means we didn't pursue this path. If someone wanted to plug TF into the ltr plugin, I still think something done via ExtensiblePlugin is probably the right way.

[1] ebernhardson@dfa7b6b

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants