I noticed in new release 1.107.3 KCPP is not recognizing the dedicated VRAM on AMD Radeon 8060S Previous versions including 1.106.2 showed it at least with approximate suggested number of GPU layers ...
If i ask my 12 or 24b model to call a tool, it doesnt stop calling it. Although it should trigger followup tool=True scenario, it seems it never does. See relevant debug output: 06:44:50] CtxLimit:236 ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results