Test now with llama.cpp

.gitignore

@@ -0,0 +1 @@
+*.swp

Dockerfile

@@ -1,13 +1,12 @@
 FROM nvidia/cuda:12.9.0-cudnn-devel-ubuntu24.04
 
 RUN DEBIAN_FRONTED=noninteractive apt update && apt upgrade -y
-RUN DEBIAN_FRONTED=noninteractive apt install -y git git-lfs build-essential cmake ninja-build curl libsdl2-dev
+RUN DEBIAN_FRONTED=noninteractive apt install -y git git-lfs build-essential cmake ninja-build curl libcurl4-openssl-dev
 
-RUN git clone https://github.com/ggml-org/whisper.cpp.git
-WORKDIR whisper.cpp
-RUN sh ./models/download-ggml-model.sh base.en
+RUN git clone https://github.com/ggml-org/llama.cpp.git
+WORKDIR llama.cpp
 ENV LD_LIBRARY_PATH="/usr/local/cuda/compat:$LD_LIBRARY_PATH"
-RUN cmake -B build -GNinja -DGGML_CUDA=1 -DWHISPER_SDL2=ON -DWHISPER_BUILD_TESTS=OFF && \
+RUN cmake -B build -GNinja -DGGML_CUDA=1 && \
     cmake --build build --config Release
 
-CMD ./build/bin/whisper-cli -m models/ggml-base.en.bin -f samples/jfk.wav --print-colors
+CMD ./build/bin/llama-server --host 0.0.0.0 -hf ggml-org/SmolVLM-256M-Instruct-GGUF

README.md

@@ -5,6 +5,7 @@ colorFrom: red
 colorTo: yellow
 sdk: docker
 pinned: false
+app_port: 8080
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference