Autotagger runtime when using more prompts

Started by bekesizl, Today at 08:53:29 AM

Previous topic - Next topic

bekesizl

It is not clear to me, how AutoTagger is running when I use multiple prompts for traits, keywords etc. 
Probably a stupid question, but it would help me planning how to run it on my library. 

The specific question is for local Gemma modell on LMStudio. 
Is Imatch sending the photo only once for all prompts?
Is LMStudio processing the prompts one at a time?

Is it quicker to process many prompts at once or rather let it run through multiple times with the special prompts. 

And is the case different with a Cloud AI like Mistral or OpenAI? 

Mario

AutoTagger uses the prompts in your setting, transforms them, combines them and sends them to the AI in a specific format. It also loads the image, transforms it into BASE64-encoded binary blob and includes that in the query it sends to the AI.
AutoTagger prompts the AI once for each file. This is the most efficient and cost-efficient (for paid AI's) mode of operation.