Update Dockerfile

Dockerfile

@@ -1,5 +1,4 @@
-# Base image
-FROM nvidia/cuda:11.8.0-devel-ubuntu22.04
+FROM nvidia/cuda:12.2.0-devel-ubuntu22.04
 
 ENV DEBIAN_FRONTEND=noninteractive
 
@@ -7,7 +6,6 @@ ENV DEBIAN_FRONTEND=noninteractive
 RUN apt update && \
     apt install --no-install-recommends -y \
         build-essential \
-        nvidia-cuda-toolkit \
         python3 \
         python3-pip \
         wget \
@@ -55,4 +53,4 @@ WORKDIR $HOME/app
 EXPOSE 8080
 
 # Start the llava-server with models
-CMD ["/app/llama.cpp/server", "--model", "ggml-model-q4_k.gguf", "--mmproj", "mmproj-model-f16.gguf", "--host", "0.0.0.0", "--threads", "4", "-ngl", "30", "-ts", "100,0"]
+CMD ["/app/llama.cpp/server", "--model", "ggml-model-q4_k.gguf", "--mmproj", "mmproj-model-f16.gguf", "--host", "0.0.0.0", "--threads", "10"]