Revamp the dynamic library shim

This switches the default llama.cpp to be CPU based, and builds the GPU variants
as dynamically loaded libraries which we can select at runtime.

This also bumps the ROCm library to version 6 given 5.7 builds don't work
on the latest ROCm library that just shipped.

llm/llama.cpp/gen_common.sh

@@ -6,7 +6,7 @@ init_vars() {
     CMAKE_DEFS="-DLLAMA_ACCELERATE=on"
     # TODO - LLAMA_K_QUANTS is stale and needs to be mapped to newer cmake settings
     CMAKE_TARGETS="--target ggml --target ggml_static --target llama --target build_info --target common --target ext_server --target llava_static"
-    if echo "${CGO_CFLAGS}" | grep -- '-g' > /dev/null ; then
+    if echo "${CGO_CFLAGS}" | grep -- '-g' >/dev/null; then
         CMAKE_DEFS="-DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_VERBOSE_MAKEFILE=on -DLLAMA_GPROF=on ${CMAKE_DEFS}"
     else
         # TODO - add additional optimization flags...
@@ -15,7 +15,7 @@ init_vars() {
 }
 
 git_module_setup() {
-    if [ -n "${OLLAMA_SKIP_PATCHING}" ] ; then
+    if [ -n "${OLLAMA_SKIP_PATCHING}" ]; then
         echo "Skipping submodule initialization"
         return
     fi
@@ -25,13 +25,13 @@ git_module_setup() {
 }
 
 apply_patches() {
-    if [ -n "${OLLAMA_SKIP_PATCHING}" ] ; then
+    if [ -n "${OLLAMA_SKIP_PATCHING}" ]; then
         echo "Skipping submodule patching"
         return
     fi
     # Workaround git apply not handling creation well for iteration
     rm -f gguf/examples/server/server.h
-    for patch in ${PATCHES} ; do
+    for patch in ${PATCHES}; do
         git -C gguf apply ../patches/${patch}
     done
 }
@@ -39,4 +39,4 @@ apply_patches() {
 build() {
     cmake -S ${LLAMACPP_DIR} -B ${BUILD_DIR} ${CMAKE_DEFS}
     cmake --build ${BUILD_DIR} ${CMAKE_TARGETS} -j8
-}
+}