Modelapi Memory Usage is High on Azure Kubernetes

Rasa Open Source
1

Hi, I have a rasa chatbot api&action released on azure kubernetes as docker images. The memory usage is super high for api (This is for processing request not for training model). How to reduce/limit the memory usage?

  • Memory usage For RASA api, the CPU is less than 1%; but the memory usage is above 91% (Usage about 1.8G, the pod is about 512 Mi, limit 1.95G) For RASA action, the CPU is less than 1% and the memory usage is about 7% (75Mi, the pod is about 512 Mi, limit 1.95G)

  • RASA version and model use rasa==3.0.8 fuzzywuzzy spacy

  • Language model use: python