llama-cli doesn't load model to GPU

[tmpuserx]

Hello, my system has a RTX3070 and CUDA installed properly. i downloaded llama-b5688-bin-win-cuda-12.4-x64 and used llama-cli to load llama-b5688-bin-win-cuda-12.4-x64 model from local. and it looks like it loaded into PC memory instead of GPU memory. anyone know how to get the model load to GPU? thanks.

[aryan8433]

Use -ngl 99

[tmpuserx]

Thanks. looks like i also have to copy the cudart-llama-bin-win-cuda-12.4-x64 runtime files into the folder so that it would work :)