I am implementing a service to analyze big documents. When query Ollama server I get a difference (of more than 5 seconds) between the "total duration" and the sum of load_duration + prompt_eval_duration + eval_duration, as seen below: {(...) total_duration":8139553772, # 8,13 Seconds "load_duration":44643610, # 0,04 Seconds "prompt_eval_count":63859, "prompt_eval_duration":205727569, # 0,20 Seconds "eval_count":82, "eval_duration":2676314808} #...
Originally asked by Marc Rosenfeld on Stack Exchange (stackoverflow) (CC BY-SA 4.0).
