layer-by-layer inference
Qwen's new dense VL (Gated DeltaNet + Gated Attention, native vision) runs in 3.33GB of VRAM, measured end to end on one RTX 3090. Needs
transformers 5.8+AirLLM
Creator
Creator
Seonglae ChoCreated
Created
2024 Oct 21 20:44Editor
Editor
Seonglae ChoEdited
Edited
2026 Aug 23 10:43Refs
Refs