Use Vulkan

Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.2.0-runtime-ubuntu22.04 AS base
+FROM nvidia/cuda:12.6.0-runtime-ubuntu24.04 AS base
 
 # Install base packages
 RUN apt-get update && \
@@ -11,7 +11,7 @@ RUN apt-get update && \
     git \
     gnupg \
     jq \
-    libcurl4-nss-dev \
+    libcurl4-gnutls-dev \
     libvulkan1 \
     libvulkan-dev \
     mesa-vulkan-drivers \
@@ -46,8 +46,6 @@ RUN mkdir -p /usr/local/lib/llama-cpp-python
 RUN tar -xf /tmp/vulkan-libs.tar -C /usr/local/lib/llama-cpp-python
 RUN cp -f /usr/local/lib/llama-cpp-python/*.so* /usr/local/lib/
 RUN rm /tmp/vulkan-libs.tar
-# Set Vulkan environment to use CPU fallback (LavaPipe)
-ENV VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/lvp_icd.x86_64.json
 
 # Create directory structure
 RUN mkdir -p \