llama.cpp docker cublas make it go fast

gistfile1.txt

docker rm -f rocm
sudo docker run -d \
    --name=rocm \
    --privileged \
    -v /dev:/dev \
    -v /mnt/persist/work/rocm:/home/rocm-user \
    --security-opt seccomp=unconfined \
    --group-add video \
    rocm/rocm-terminal \
    bash -c "sleep infinity"

docker exec --user 0 -it rocm bash -c "mkdir -p /mnt/models && mount /dev/disk/by-partlabel/EXT_STORAGE /mnt/models/"
docker exec -it rocm bash

sudo apt update
sudo apt install pkg-config build-essential

git clone --recurse-submodules https://github.com/KhronosGroup/OpenCL-SDK.git
mkdir -p OpenCL-SDK/build
cd OpenCL-SDK/build
cmake .. -DBUILD_DOCS=OFF \
  -DBUILD_EXAMPLES=OFF \
  -DBUILD_TESTING=OFF \
  -DOPENCL_SDK_BUILD_SAMPLES=OFF \
  -DOPENCL_SDK_TEST_SAMPLES=OFF
cmake --build . --config Release -j8
sudo cmake --install . --prefix /usr/local

git clone https://github.com/CNugteren/CLBlast.git
mkdir -p CLBlast/build
cd CLBlast/build
cmake .. -DBUILD_SHARED_LIBS=OFF -DTUNERS=OFF -DCMAKE_POSITION_INDEPENDENT_CODE=ON
cmake --build . --config Release -j8
sudo cmake --install . --prefix /usr/local

git clone https://github.com/ggerganov/llama.cpp -b dadbed99e65252d79f81101a392d0d6497b86caa
make LLAMA_CLBLAST=1 -j8

rm models || true
ln -fs /mnt/models/backup/models/llama ./models

Remove --threads option
Change num gpu layers to 42 (or whatever the maximum is)