Tensor split mode: CUDA error on latest llama.cpp with Qwen-3.6-27b

Article automatically generated from technical news.

Hi guys, I am running into issues when loading the Unsloth UD-Q8_K_XL quant and wanted to check if anyone has ran into this. I updated my config to also use --split-mode tensor but wanted to check if I need to update drivers/CUDA to get it working as I see that the tensor split mode fixes are merged into llama.cpp. Running dual 3090's on Ubuntu Server 24.04. NVIDIA-SMI 580.159.03 Driver Version: 580.159.03 CUDA Version: 13.0 This is my config running in Docker with the latest llama.cp