how to accelerate the response time？often take 2-3 seconds,how to reduce to 1 seconds.Except use ubuntu server，and how to write some codes in rasa to accelerate response time?
Hi, Can you share more.info like -
- Rasa version
- How many intents do you have ?
- How many entities
Usually the responses I have seen are in milliseconds.
rasa3.0; have 450 intents and 7000 rows;entities have 11 kinds
Hi @xhyxxj, 450 intents sound really a lot!. Maybe using retrieval intents is an option. This way you can group your intent and first only the group is predicted and second the specific intent response key. This can also drastically reduce the rules you need for your bot: Chitchat and FAQs
- use a GPU
- optimize custom action code
- use a simpler pipeline and policy configuration