Merge pull request #369 from containers/fedora-asahi

Start an Asahi version

README.md

@@ -92,6 +92,7 @@ curl -fsSL https://raw.githubusercontent.com/containers/ramalama/s/install.sh |
 | Hardware                           | Enabled |
 | ---------------------------------- | ------- |
 | CPU                                | :white_check_mark: |
+| Apple Silicon GPU (Linux / Asahi)  | :white_check_mark: |
 | Apple Silicon GPU (macOS)          | :white_check_mark: |
 | Apple Silicon GPU (podman-machine) | :x: |
 | Nvidia GPU (cuda)                  | :x: [Containerfile](https://github.com/containers/ramalama/blob/main/container-images/cuda/Containerfile) available but not published to quay.io |

container-images/asahi/Containerfile

@@ -0,0 +1,25 @@
+FROM fedora:41
+
+# renovate: datasource=github-releases depName=containers/omlmd extractVersion=^v(?<version>.*)
+ARG OMLMD_VERSION=0.1.6
+ARG LLAMA_CPP_SHA=1329c0a75e6a7defc5c380eaf80d8e0f66d7da78
+# renovate: datasource=git-refs depName=ggerganov/whisper.cpp packageName=https://github.com/ggerganov/whisper.cpp gitRef=master versioning=loose type=digest
+ARG WHISPER_CPP_SHA=0377596b77a3602e36430320cbe45f8c305ef04a
+
+RUN dnf install -y dnf-plugins-core && \
+    dnf copr enable -y @asahi/fedora-remix-branding && \
+    dnf install -y asahi-repos && \
+    dnf install -y mesa-vulkan-drivers vulkan-headers vulkan-loader-devel \
+      vulkan-tools spirv-tools glslc glslang git procps-ng vim cmake gcc-c++ \
+      python3-pip python3-argcomplete && \
+    dnf clean all && \
+    rm -rf /var/cache/*dnf*
+
+RUN /usr/bin/python3 --version
+RUN pip install "omlmd==${OMLMD_VERSION}"
+
+COPY ../scripts /scripts
+RUN chmod +x /scripts/*.sh && \
+    /scripts/build_llama_and_whisper.sh "$LLAMA_CPP_SHA" "$WHISPER_CPP_SHA" \
+      "/usr" "-DGGML_VULKAN=1"
+

container-images/cuda/Containerfile

@@ -10,7 +10,7 @@ RUN dnf install -y git cmake gcc-c++ && \
     dnf clean all && rm -rf /var/cache/*dnf*
 
 # Set the temporary installation directory
-ENV INSTALL_PREFIX=/tmp/install
+ARG INSTALL_PREFIX=/tmp/install
 
 COPY ../scripts /scripts
 RUN chmod +x /scripts/*.sh && \
@@ -21,8 +21,6 @@ RUN chmod +x /scripts/*.sh && \
 # Final runtime image
 FROM docker.io/nvidia/cuda:12.6.2-runtime-ubi9
 
-# renovate: datasource=github-releases depName=huggingface/huggingface_hub extractVersion=^v(?<version>.*)
-ARG HUGGINGFACE_HUB_VERSION=0.26.2
 # renovate: datasource=github-releases depName=containers/omlmd extractVersion=^v(?<version>.*)
 ARG OMLMD_VERSION=0.1.6
 
@@ -32,8 +30,7 @@ RUN dnf install -y python3 python3-pip && \
     rm -rf /var/cache/*dnf*
 
 # Install Python packages in the runtime image
-RUN pip install "huggingface_hub==${HUGGINGFACE_HUB_VERSION}" \
-      "omlmd==${OMLMD_VERSION}"
+RUN pip install "omlmd==${OMLMD_VERSION}"
 
 # Copy the entire installation directory from the builder
 COPY --from=builder /tmp/install /usr

container-images/ramalama/Containerfile

@@ -1,11 +1,7 @@
 FROM registry.access.redhat.com/ubi9/ubi:9.4-1214.1729773476
 
-# renovate: datasource=github-releases depName=huggingface/huggingface_hub extractVersion=^v(?<version>.*)
-ARG HUGGINGFACE_HUB_VERSION=0.26.2
 # renovate: datasource=github-releases depName=containers/omlmd extractVersion=^v(?<version>.*)
 ARG OMLMD_VERSION=0.1.6
-# renovate: datasource=github-releases depName=tqdm/tqdm extractVersion=^v(?<version>.*)
-ARG TQDM_VERSION=4.66.6
 ARG LLAMA_CPP_SHA=1329c0a75e6a7defc5c380eaf80d8e0f66d7da78
 # renovate: datasource=git-refs depName=ggerganov/whisper.cpp packageName=https://github.com/ggerganov/whisper.cpp gitRef=master versioning=loose type=digest
 ARG WHISPER_CPP_SHA=0377596b77a3602e36430320cbe45f8c305ef04a
@@ -24,9 +20,7 @@ RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.n
     rm -rf /var/cache/*dnf*
 
 RUN /usr/bin/python3 --version
-RUN pip install "huggingface_hub==${HUGGINGFACE_HUB_VERSION}"
 RUN pip install "omlmd==${OMLMD_VERSION}"
-RUN pip install "tqdm==${TQDM_VERSION}"
 
 RUN dnf config-manager --add-repo \
       https://mirror.stream.centos.org/9-stream/AppStream/$(uname -m)/os/

ramalama/model.py

@@ -105,10 +105,16 @@ def model_path(self, args):
         raise NotImplementedError(f"model_path for {self.type} not implemented")
 
     def _image(self, args):
+        if args.image != default_image():
+            return args.image
+
         gpu_type, _ = get_gpu()
         if gpu_type == "HIP_VISIBLE_DEVICES":
-            if args.image == default_image():
-                return "quay.io/ramalama/rocm:latest"
+            return "quay.io/ramalama/rocm:latest"
+
+        if gpu_type == "ASAHI_VISIBLE_DEVICES":
+            return "quay.io/ramalama/asahi:latest"
+
         return args.image
 
     def setup_container(self, args):
@@ -149,7 +155,7 @@ def setup_container(self, args):
             conman_args += ["--device", "/dev/kfd"]
 
         gpu_type, gpu_num = get_gpu()
-        if gpu_type == "HIP_VISIBLE_DEVICES":
+        if gpu_type == "HIP_VISIBLE_DEVICES" or gpu_type == "ASAHI_VISIBLE_DEVICES":
             conman_args += ["-e", f"{gpu_type}={gpu_num}"]
         return conman_args
 
@@ -200,7 +206,7 @@ def gpu_args(self):
             # any additional arguments.
             pass
         elif sys.platform == "linux" and (
-            os.path.exists("/dev/dri") or os.getenv("HIP_VISIBLE_DEVICES") or os.getenv("CUDA_VISIBLE_DEVICES")
+            os.getenv("HIP_VISIBLE_DEVICES") or os.getenv("ASAHI_VISIBLE_DEVICES") or os.getenv("CUDA_VISIBLE_DEVICES")
         ):
             gpu_args = ["-ngl", "99"]
         else:
@@ -335,6 +341,12 @@ def get_gpu():
     if gpu_bytes:  # this is the ROCm/AMD case
         return "HIP_VISIBLE_DEVICES", gpu_num
 
+    if os.path.exists('/etc/os-release'):
+        with open('/etc/os-release', 'r') as file:
+            content = file.read()
+            if "asahi" in content.lower():
+                return "ASAHI_VISIBLE_DEVICES", 1
+
     return None, None
 
 

test/ci.sh

@@ -32,6 +32,8 @@ main() {
   # verify llama.cpp version matches
   grep "$(grep "ARG LLAMA_CPP_SHA=" container-images/ramalama/Containerfile)" \
     container-images/cuda/Containerfile
+  grep "$(grep "ARG LLAMA_CPP_SHA=" container-images/ramalama/Containerfile)" \
+    container-images/asahi/Containerfile
 
   local os
   os="$(uname -s)"