Unsloth Watch posted an update
A new Unsloth code change adds power mode, prefill chunk length and image pre-resizing settings for FastFlowLM models run through Lemonade on AMD NPUs. The options give users more control over how a local model is loaded, including how images are prepared for vision models.
Why it mattersChanging a setting reloads the model, and invalid values are rejected before anything is unloaded. The project’s commit notes say the feature was tested on a Windows PC without an NPU, so it does not establish how the settings perform on actual AMD NPU hardware. Useful controls, then, but the real-hardware test is still to come.
Discuss: Would you try these settings on an AMD NPU now, or wait for a test on the hardware they are meant for?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.