News for hackers. Stuff that matters.
Re:Qwen2.5-VL-32B: Smarter and Lighter by Anaphylaxis on Tuesday March 25, 2025 @07:31AM (#43470006)
Almost guaranteed this is user error. Ollama has a (tiny) default of 2048 context, so about then is probably when you noticed the results sharply decline in quality. Try 16384

There are no comments on this comment yet.