diff --git a/Dockerfile.build b/Dockerfile.build
index f1f5f3fc..5499b0a1 100644
--- a/Dockerfile.build
+++ b/Dockerfile.build
@@ -1,11 +1,18 @@
 # Ubuntu 20.04 amd64 dependencies
-FROM --platform=linux/amd64 nvidia/cuda:11.7.1-devel-ubuntu22.04 AS base-amd64
+FROM --platform=linux/amd64 ubuntu:20.04 AS base-amd64
+ARG CUDA_VERSION=11.3.1-1
+ARG CMAKE_VERSION=3.22.1
 # ROCm only supports amd64
 ARG ROCM_VERSION=5.7
+
 # Note: https://rocm.docs.amd.com/en/latest/release/user_kernel_space_compat_matrix.html
 RUN apt-get update && \
-    apt-get install -y wget && \
-    wget "https://github.com/Kitware/CMake/releases/download/v3.22.1/cmake-3.22.1-linux-x86_64.sh" -O /tmp/cmake-installer.sh && \
+    apt-get install -y wget gnupg && \
+    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin && \
+    mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600 && \
+    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub && \
+    echo "deb [by-hash=no] https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/ /" > /etc/apt/sources.list.d/cuda.list && \
+    wget "https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-x86_64.sh" -O /tmp/cmake-installer.sh && \
     chmod +x /tmp/cmake-installer.sh && /tmp/cmake-installer.sh --skip-license --prefix=/usr && \
     mkdir --parents --mode=0755 /etc/apt/keyrings && \
     wget https://repo.radeon.com/rocm/rocm.gpg.key -O - | gpg --dearmor > /etc/apt/keyrings/rocm.gpg && \
@@ -14,35 +21,45 @@ RUN apt-get update && \
     echo "Pin: release o=repo.radeon.com" >> /etc/apt/preferences.d/rocm-pin-600 && \
     echo "Pin-Priority: 600" >> /etc/apt/preferences.d/rocm-pin-600 && \
     apt-get update && \
-    DEBIAN_FRONTEND=noninteractive apt-get install -y rocm-hip-libraries rocm-device-libs rocm-libs rocm-ocl-icd rocm-hip-sdk rocm-hip-libraries rocm-cmake rocm-clang-ocl rocm-dev
+    DEBIAN_FRONTEND=noninteractive apt-get -y install cuda=${CUDA_VERSION} rocm-hip-libraries rocm-device-libs rocm-libs rocm-ocl-icd rocm-hip-sdk rocm-hip-libraries rocm-cmake rocm-clang-ocl rocm-dev
 
 ENV ROCM_PATH=/opt/rocm
 
 # Ubuntu 22.04 arm64 dependencies
-FROM --platform=linux/arm64 nvidia/cuda:11.7.1-devel-ubuntu22.04 AS base-arm64
+FROM --platform=linux/arm64 ubuntu:20.04 AS base-arm64
+ARG CUDA_VERSION=11.3.1-1
+ARG CMAKE_VERSION=3.27.6
 RUN apt-get update && \
-    apt-get install -y wget && \
-    wget "https://github.com/Kitware/CMake/releases/download/v3.27.6/cmake-3.27.6-linux-aarch64.sh" -O /tmp/cmake-installer.sh && \
-    chmod +x /tmp/cmake-installer.sh && /tmp/cmake-installer.sh --skip-license --prefix=/usr
+    apt-get install -y wget gnupg && \
+    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/sbsa/cuda-ubuntu2004.pin && \
+    mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600 && \
+    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/sbsa//3bf863cc.pub && \
+    echo "deb [by-hash=no] https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/sbsa/ /" > /etc/apt/sources.list.d/cuda.list && \
+    wget "https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-aarch64.sh" -O /tmp/cmake-installer.sh && \
+    chmod +x /tmp/cmake-installer.sh && /tmp/cmake-installer.sh --skip-license --prefix=/usr && \
+    apt-get update && \
+    apt-cache madison cuda && \
+    DEBIAN_FRONTEND=noninteractive apt-get -y install cuda=${CUDA_VERSION} 
 
 FROM base-${TARGETARCH}
 ARG TARGETARCH
 ARG GOFLAGS="'-ldflags -w -s'"
 ARG CGO_CFLAGS
 ARG CLBLAST_VER=1.6.1
+ARG GOLANG_VERSION=1.21.3
 
 # Common toolchain
 RUN apt-get update && \
-    DEBIAN_FRONTEND=noninteractive apt-get install -y gcc-11 g++-11 cpp-11 git ocl-icd-opencl-dev && \
-    update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 100 --slave /usr/bin/g++ g++ /usr/bin/g++-11 --slave /usr/bin/gcov gcov /usr/bin/gcov-11
+    DEBIAN_FRONTEND=noninteractive apt-get install -y gcc-10 g++-10 cpp-10 git ocl-icd-opencl-dev && \
+    update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-10 100 --slave /usr/bin/g++ g++ /usr/bin/g++-10 --slave /usr/bin/gcov gcov /usr/bin/gcov-10
 
 # CLBlast
 RUN wget -qO- https://github.com/CNugteren/CLBlast/archive/refs/tags/${CLBLAST_VER}.tar.gz | tar zxv -C /tmp/ && \
     cd /tmp/CLBlast-${CLBLAST_VER} && mkdir build && cd build && cmake .. && make && make install
 
 # install go
-ADD https://dl.google.com/go/go1.21.3.linux-$TARGETARCH.tar.gz /tmp/go1.21.3.tar.gz
-RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.3.tar.gz
+ADD https://dl.google.com/go/go${GOLANG_VERSION}.linux-$TARGETARCH.tar.gz /tmp/go${GOLANG_VERSION}.tar.gz
+RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go${GOLANG_VERSION}.tar.gz
 
 # build the final binary
 WORKDIR /go/src/github.com/jmorganca/ollama
diff --git a/llm/llama.cpp/gen_common.sh b/llm/llama.cpp/gen_common.sh
index 2f75104f..83a21cf9 100644
--- a/llm/llama.cpp/gen_common.sh
+++ b/llm/llama.cpp/gen_common.sh
@@ -15,13 +15,20 @@ init_vars() {
 }
 
 git_module_setup() {
-    # TODO add flags to skip the init/patch logic to make it easier to mod llama.cpp code in-repo
+    if [ -n "${OLLAMA_SKIP_PATCHING}" ] ; then
+        echo "Skipping submodule initialization"
+        return
+    fi
     git submodule init
     git submodule update --force gguf
 
 }
 
 apply_patches() {
+    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
diff --git a/llm/rocm_shim.c b/llm/rocm_shim.c
index 9a6595b1..e8304aa0 100644
--- a/llm/rocm_shim.c
+++ b/llm/rocm_shim.c
@@ -3,13 +3,13 @@
 #include <stdio.h>
 #include <string.h>
 
-#ifndef _WIN32
+#ifdef __linux__
 #include <dlfcn.h>
-#define LOAD_LIBRARY(lib, flags) dlopen(lib, flags)
+#define LOAD_LIBRARY(lib, flags) dlopen(lib, flags | RTLD_DEEPBIND)
 #define LOAD_SYMBOL(handle, sym) dlsym(handle, sym)
 #define LOAD_ERR() dlerror()
 #define UNLOAD_LIBRARY(handle) dlclose(handle)
-#else
+#elif _WIN32
 #include <windows.h>
 #define LOAD_LIBRARY(lib, flags) LoadLibrary(lib)
 #define LOAD_SYMBOL(handle, sym) GetProcAddress(handle, sym)
@@ -20,6 +20,12 @@ inline static char *LOAD_ERR() {
   snprintf(errbuf, 8, "0x%lx", GetLastError());
   return errbuf;
 }
+#else
+#include <dlfcn.h>
+#define LOAD_LIBRARY(lib, flags) dlopen(lib, flags)
+#define LOAD_SYMBOL(handle, sym) dlsym(handle, sym)
+#define LOAD_ERR() dlerror()
+#define UNLOAD_LIBRARY(handle) dlclose(handle)
 #endif
 
 void rocm_shim_init(const char *libPath, struct rocm_llama_server *s,
@@ -48,7 +54,7 @@ void rocm_shim_init(const char *libPath, struct rocm_llama_server *s,
   };
 
   printf("Lazy loading %s library\n", libPath);
-  s->handle = LOAD_LIBRARY(libPath, RTLD_LAZY);
+  s->handle = LOAD_LIBRARY(libPath, RTLD_NOW);
   if (!s->handle) {
     err->id = -1;
     snprintf(