Dockerfile

FROM ghcr.io/ggml-org/llama.cpp:full

RUN apt update && apt install wget -y

RUN wget "https://huggingface.co/unsloth/gemma-3-270m-it-GGUF/resolve/main/gemma-3-270m-it-F16.gguf" -O /gemma-3-270m-it-F16.gguf

# Optimized for Gemma 3's native 32k context window.
# WARNING: Hardware limitations (RAM/CPU) are the main concern.
CMD ["--server", "-m", "/gemma-3-270m-it-F16.gguf", "--port", "7860", "--host", "0.0.0.0", "-c", "32000", "-n", "4000", "-t", "2", "--mlock"]