ggml-cuda.cu:3211: ERROR: CUDA kernel vec_dot_q5_K_q8_1_impl_vmmq

[DanCard]

Discussed in #5685

^{Originally posted by DanCard February 23, 2024}
ggml-cuda.cu:3211: ERROR: CUDA kernel vec_dot_q5_K_q8_1_impl_vmmq has no device code compatible with CUDA arch 520. ggml-cuda.cu was compiled for: 520
This worked yesterday. I did a git pull, make clean, make and then get this error today.
nvidia rtx 3090
System: debian testing
Command line:
~/github/llama.cpp/main -m ~/models/miqu-1-70b.q5_K_M.gguf -c 0 -i --color -t 16 --n-gpu-layers 24 --temp 0.8 -p "bob"

I reverted previous two commits and issue went away.
~/github/llama.cpp$ git reset --hard HEAD~2
HEAD is now at 334f76f sync : ggml

[DanCard]

This is no longer an issue with latest update.

[szymonrucinski]

Hey I still keep getting this error for the server. I am on the main branch as well.