From 96a8ea024102b94a9ff00f2044860f619cc4abe5 Mon Sep 17 00:00:00 2001
From: Benson Wong <mostlygeek@gmail.com>
Date: Fri, 14 Feb 2025 15:25:45 -0800
Subject: [PATCH] add cpu docker container build

---
 README.md                 |  7 +++++--
 docker/build-container.sh | 24 +++++++++++++++++++-----
 2 files changed, 24 insertions(+), 7 deletions(-)

diff --git a/README.md b/README.md
index 1b13e30..e42de16 100644
--- a/README.md
+++ b/README.md
@@ -31,7 +31,8 @@ Written in golang, it is very easy to install (single binary with no dependancie
 Docker is the quickest way to try out llama-swap:
 
 ```
-$ docker run -it --rm --runtime nvidia -p 9292:8080 ghcr.io/mostlygeek/llama-swap:cuda
+# use CPU inference
+$ docker run -it --rm -p 9292:8080 ghcr.io/mostlygeek/llama-swap:cpu
 
 
 # qwen2.5 0.5B
@@ -52,10 +53,12 @@ $ curl -s http://localhost:9292/v1/chat/completions \
 
 Docker images are [published nightly](https://github.com/mostlygeek/llama-swap/pkgs/container/llama-swap) that include the latest llama-swap and llama-server:
 
+- `ghcr.io/mostlygeek/llama-swap:cpu`
 - `ghcr.io/mostlygeek/llama-swap:cuda`
 - `ghcr.io/mostlygeek/llama-swap:intel`
 - `ghcr.io/mostlygeek/llama-swap:vulkan`
-- `ghcr.io/mostlygeek/llama-swap:musa`
+- ROCm disabled until fixed in llama.cpp container
+- musa disabled until requested.
 
 Specific versions are also available and are tagged with the llama-swap, architecture and llama.cpp versions. For example: `ghcr.io/mostlygeek/llama-swap:v89-cuda-b4716`
 
diff --git a/docker/build-container.sh b/docker/build-container.sh
index 97839d7..90b39a9 100755
--- a/docker/build-container.sh
+++ b/docker/build-container.sh
@@ -5,7 +5,7 @@ cd $(dirname "$0")
 ARCH=$1
 
 # List of allowed architectures
-ALLOWED_ARCHS=("intel" "vulkan" "musa" "cuda")
+ALLOWED_ARCHS=("intel" "vulkan" "musa" "cuda", "cpu")
 
 # Check if ARCH is in the allowed list
 if [[ ! " ${ALLOWED_ARCHS[@]} " =~ " ${ARCH} " ]]; then
@@ -22,10 +22,24 @@ fi
 # the most recent llama-swap tag
 # have to strip out the 'v' due to .tar.gz file naming
 LS_VER=$(curl -s https://api.github.com/repos/mostlygeek/llama-swap/releases/latest | jq -r .tag_name | sed 's/v//')
-LCPP_TAG=$(curl -s -H "Authorization: Bearer $GITHUB_TOKEN" \
-    "https://api.github.com/users/ggerganov/packages/container/llama.cpp/versions" \
-    | jq -r --arg arch "$ARCH" '.[] | select(.metadata.container.tags[] | startswith("server-\($arch)")) | .metadata.container.tags[]' \
-    | sort -r | head -n1 | awk -F '-' '{print $3}')
+
+if [ "$ARCH" == "cpu" ]; then
+    # cpu only containers just use the latest available
+    CONTAINER_LATEST="ghcr.io/mostlygeek/llama-swap:cpu"
+    echo "Building ${CONTAINER_LATEST} $LS_VER"
+    docker build -f llama-swap.Containerfile --build-arg BASE_TAG=server --build-arg LS_VER=${LS_VER} -t ${CONTAINER_LATEST} .
+else
+    LCPP_TAG=$(curl -s -H "Authorization: Bearer $GITHUB_TOKEN" \
+        "https://api.github.com/users/ggerganov/packages/container/llama.cpp/versions" \
+        | jq -r --arg arch "$ARCH" '.[] | select(.metadata.container.tags[] | startswith("server-\($arch)")) | .metadata.container.tags[]' \
+        | sort -r | head -n1 | awk -F '-' '{print $3}')
+
+    CONTAINER_TAG="ghcr.io/mostlygeek/llama-swap:v${LS_VER}-${ARCH}-${LCPP_TAG}"
+    CONTAINER_LATEST="ghcr.io/mostlygeek/llama-swap:${ARCH}"
+    echo "Building ${CONTAINER_TAG} $LS_VER"
+    docker build -f llama-swap.Containerfile --build-arg BASE_TAG=server-${ARCH}-${LCPP_TAG} --build-arg LS_VER=${LS_VER} -t ${CONTAINER_TAG} -t ${CONTAINER_LATEST} .
+fi
+exit
 
 CONTAINER_TAG="ghcr.io/mostlygeek/llama-swap:v${LS_VER}-${ARCH}-${LCPP_TAG}"
 CONTAINER_LATEST="ghcr.io/mostlygeek/llama-swap:${ARCH}"