search

facebookresearch / ReAgent

A platform for Reasoning systems (Reinforcement Learning, Contextual Bandits, etc.)
https://reagent.ai
BSD 3-Clause "New" or "Revised" License
3.57k stars 521 forks source link

Where to find examples on how to consume the model? #324

Open susanhuhu opened 4 years ago

susanhuhu commented 4 years ago

Is there a code example on how to consume the model? And does the model only support single slot decision making? Meaning the input will be a list of possible actions + context, and then caller can call the model to get one suggested action. Is there support for multiple slots decision like CCB described here https://github.com/VowpalWabbit/vowpal_wabbit/wiki/Conditional-Contextual-Bandit?

MisterTea commented 4 years ago

Hey @susanhuhu,

This tutorial has an example of calling the model: https://reagent.ai/rasp_tutorial.html

We have several models for ranking but they are new and the documentation is still pretty sparse https://github.com/facebookresearch/ReAgent/tree/master/reagent/training/ranking