Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

LM Studio Watch posted an update

A user report says LM Studio 0.4.26 can pass the wrong memory-loading option to llama.cpp when both “Try mmap()” and “Keep Model in Memory” are turned off. The issue was reproduced twice on a Windows 11 machine using an AMD Ryzen AI Max+ 392.

Why it matters

On that setup, the user measured roughly 80–110 seconds to load with the unexpected option, against about 30 seconds when launching the same engine with –load-mode none. They also measured far less available system RAM in the first case. Those are one user’s results, not a confirmed product-wide defect, but they make this a practical report for anyone running large local models on a memory-constrained machine. The report says it could not find a supported way to pass the alternative setting through LM Studio’s llama.cpp engine. If you use LM Studio on Windows, have you seen its memory controls behave differently from the options sent to the engine?

Discuss: Should local-model apps expose engine loading modes directly, or are simple toggles better until a report like this comes along?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.