Use CUDA llama.cpp server image

Dockerfile

@@ -1,17 +1,17 @@
-FROM ghcr.io/ggml-org/llama.cpp:server
+FROM ghcr.io/ggml-org/llama.cpp:server-cuda
 
 ENV MODEL_REPO=sayshara/nemotron-3-nano-4b-ascii-art-GGUF
 ENV MODEL_FILE=nemotron-3-nano-4b-ascii-art-q4_k_m.gguf
 ENV PORT=7860
 ENV CTX_SIZE=2048
-ENV N_GPU_LAYERS=0
-ENV HF_HUB_ENABLE_HF_TRANSFER=1
+ENV N_GPU_LAYERS=99
+ENV HF_XET_HIGH_PERFORMANCE=1
 
 USER root
 RUN apt-get update \
     && apt-get install -y --no-install-recommends python3 python3-pip ca-certificates \
     && rm -rf /var/lib/apt/lists/* \
-    && python3 -m pip install --break-system-packages --no-cache-dir "huggingface_hub[hf_transfer]"
+    && python3 -m pip install --break-system-packages --no-cache-dir "huggingface_hub"
 
 COPY start_server.py /app/start_server.py