The Slow Reply Was CPU Time
The Slow Reply Was CPU Time Thirteen gigabytes of memory were still free, and the answer was still slow. Savindu Thathsara, savithathsara.me . This is a Hearthlen note from the laptop. Sinhala had just been added beside English, and a question took long enough to feel stuck. The check was /proc/meminfo : 13.0 GB available out of 15.5 GB. The listener was 1507 MB . The language model was not…
Thirteen gigabytes of memory remained free on the laptop despite a slow response time, as reported in a note from Savindu Thathsara. The model was not overloaded with memory, but the sluggishness stemmed from the language model's work. When a question was asked, the bilingual speech model spent 2407 milliseconds to process the query, as it had to simultaneously predict the language.
The Sinhala reply from Qwen2.5-3B model took an even longer 9564 milliseconds. In contrast, the English model processed the response much quicker. After a restart, the listener memory usage dropped back to 1331 MB, while 13.0 GB remained unused. The 3B model remains the primary thinker, generating replies that still took a few seconds. Savindu Thathsara, based in London, specializes in crafting practical AI and automation solutions.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
