run main: CUDA error 222

[vicwer]

run ./main -m ./main -m .s/llama-13b/ggml-model-q4_0.bin, error:

main: build = 583 (7e4ea5b)
main: seed  = 1685433863
CUDA error 222 at .../llama.cpp/ggml-cuda.cu:420: the provided PTX was compiled with an unsupported toolchain.

sdk version:

llama.cpp 7e4ea5beff567f53be92f75f9089e6f11fa5dabd
python==3.9.11
nvcc==cuda_11.8.r11.8/compiler.31833905_0
Driver==515.86.01

[vicwer]

Done: Driver==515.86.01 match nvcc==Build cuda_11.7.r11.7/compiler.31294372_0