Added pre-built images

Dockerfile.vulkan

@@ -5,7 +5,7 @@ RUN dnf install -y \
       git vim \
       make gcc cmake ninja-build lld clang clang-devel compiler-rt libcurl-devel \
       vulkan-loader-devel mesa-vulkan-drivers \
-      glslc \
+      radeontop glslc \
     && dnf clean all
 
 WORKDIR /opt/llama.cpp

README.md

@@ -2,8 +2,8 @@
 
 This repository provides two Fedora Rawhide–based Dockerfiles for building isolated dev containers with AMD’s new **Strix Halo** GPUs (gfx1151):
 
-- **Docker.rocm**: Builds `llama.cpp` with ROCm (HIP) support targeting gfx1151
+- **Dockerfile.rocm**: Builds `llama.cpp` with ROCm (HIP) support targeting gfx1151
-- **Docker.vulkan**: Builds `llama.cpp` with Vulkan compute support
+- **Dockerfile.vulkan**: Builds `llama.cpp` with Vulkan compute support
 
 Both images load the latest ROCm/Vulkan libraries from Fedora Rawhide to ensure compatibility with Strix Halo.
 
@@ -11,33 +11,78 @@ Both images load the latest ROCm/Vulkan libraries from Fedora Rawhide to ensure
 
 ```
 .
-├── Docker.rocm        # HIP-based build for ROCm (gfx1151)
+├── Dockerfile.rocm        # HIP-based build for ROCm (gfx1151)
-├── Docker.vulkan      # Vulkan-based build
+├── Dockerfile.vulkan      # Vulkan-based build
 └── README.md              # This documentation
 ```
 
 ## Prerequisites
 
 - **Podman** (or Docker aliased to Podman)
-- **Toolbx** (for creating interactive dev containers, https://containertoolbx.org/)
+- **Toolbox** (for creating interactive dev containers, https://containertoolbx.org/)
 - Recent Linux kernel with AMD GPU drivers (`amdgpu`) installed on the host
 
+## 1. Pulling Pre-built Containers
+
+If you don't want to build locally, pull pre-built images from the registry:
+
+```bash
+podman pull docker.io/kyuz0/amd-strix-halo-toolboxes:rocm
+podman pull docker.io/kyuz0/amd-strix-halo-toolboxes:vulkan
+```
+
+```bash
+toolbox create llama-rocm \
+  --image docker.io/kyuz0/amd-strix-halo-toolboxes:rocm \
+  -- \
+    --device /dev/kfd \
+    --device /dev/dri \
+    --group-add video \
+    --security-opt seccomp=unconfined
+```
+
+```bash
+toolbox create llama-vulkan \
+  --image docker.io/kyuz0/amd-strix-halo-toolboxes:vulkan \
+  -- \
+    --device /dev/dri \
+    --group-add video \
+    --security-opt seccomp=unconfined
+```
+
+## 2. Building the Images
+
+### ROCm (HIP) Image
+
+```bash
+podman build -t llama-rocm -f Dockerfile.rocm .
+```
+
+### Vulkan Image
+
+```bash
+podman build -t llama-vulkan -f Dockerfile.vulkan .
+```
+
+
+- **Podman** (or Docker aliased to Podman)
+- **Toolbox** (for creating interactive dev containers, https://containertoolbx.org/)
+- Recent Linux kernel with AMD GPU drivers (`amdgpu`) installed on the host
 
 ## 1. Building the Images
 
 ### ROCm (HIP) Image
 
 ```bash
-podman build -t llama-rocm -f Docker.rocm .
+podman build -t llama-rocm -f Dockerfile.rocm .
 ```
 
 ### Vulkan Image
 
 ```bash
-podman build -t llama-vulkan -f Docker.vulkan .
+podman build -t llama-vulkan -f Dockerfile.vulkan .
 ```
 
-
 ## 2. Creating Toolbox Containers
 
 Toolbox will automatically mount your home directory, map your UID:GID, enable X11, and use the host network. You only need to pass through the GPU devices and relax seccomp.
@@ -46,7 +91,7 @@ Toolbox will automatically mount your home directory, map your UID:GID, enable X
 
 ```bash
 toolbox create llama-rocm \
-  --image localhost/llama-rocm:latest \
+  --image docker.io/kyuz0/amd-strix-halo-toolboxes:rocm \
   -- \
     --device /dev/kfd \
     --device /dev/dri \
@@ -54,18 +99,19 @@ toolbox create llama-rocm \
     --security-opt seccomp=unconfined
 ```
 
+> **Note:** The `--` separator tells `toolbox` to pass subsequent flags to the underlying Podman/Docker command.
+
 ### Vulkan Toolbox
 
 ```bash
 toolbox create llama-vulkan \
-  --image localhost/llama-vulkan:latest \
+  --image docker.io/kyuz0/amd-strix-halo-toolboxes:vulkan \
   -- \
     --device /dev/dri \
     --group-add video \
     --security-opt seccomp=unconfined
 ```
 
-
 ## 3. Entering and Testing
 
 After creation, enter each container and verify that your GPU and libraries are accessible.
@@ -86,12 +132,11 @@ You should see your Strix Halo gfx1151 device listed.
 toolbox enter llama-vulkan
 # inside container:
 vulkaninfo | head -n 10
-llama-cli --help  
+llama-cli --help  # run the Vulkan-enabled CLI
 ```
 
 If `vulkaninfo` reports your GPU and `llama-cli` runs without errors, the Vulkan build is working.
 
-
 ## Host Configuration
 
 The following host details and kernel settings ensure optimal performance and unified memory access for Strix Halo:
@@ -114,5 +159,3 @@ The following host details and kernel settings ensure optimal performance and un
 ## Notes
 
 Both images pull Fedora Rawhide packages for the newest ROCm/Vulkan support.
-
-