Dockerfile

FROM debian:stable-slim

ENV DEBIAN_FRONTEND=noninteractive

# passed from space environment
ARG MODEL_ID="unsloth/Qwen3-1.7B-GGUF"
ARG QUANT="Q4_K_M"
ARG SERVED_NAME="qwen-nano"
ARG PARALLEL=4
ARG CTX_SIZE=8192
ARG KV_CACHE_QUANT="q8_0"
ARG EMBEDDING_ONLY=0
ARG RERANK_ONLY=0

# llama.cpp env configs
ENV LLAMA_ARG_HF_REPO="${MODEL_ID}:${QUANT}"
ENV LLAMA_ARG_CTX_SIZE=${CTX_SIZE}
ENV LLAMA_ARG_BATCH=512
ENV LLAMA_ARG_N_PARALLEL=${PARALLEL}
ENV LLAMA_ARG_FLASH_ATTN=1
ENV LLAMA_ARG_CACHE_TYPE_K="${KV_CACHE_QUANT}"
ENV LLAMA_ARG_CACHE_TYPE_V="${KV_CACHE_QUANT}"
ENV LLAMA_ARG_MLOCK=1
ENV LLAMA_ARG_N_GPU_LAYERS=0
ENV LLAMA_ARG_HOST="0.0.0.0"
ENV LLAMA_ARG_PORT=7860
ENV LLAMA_ARG_ALIAS="${SERVED_NAME}"
ENV LLAMA_ARG_EMBEDDINGS=${EMBEDDING_ONLY}
ENV LLAMA_ARG_RERANKING=${RERANK_ONLY}

RUN apt-get update && \
    apt-get install -y --no-install-recommends \
    git \
    curl \
    cmake \
    ccache \
    pkg-config \
    build-essential \
    ca-certificates  \
    libboost-system-dev \
    libcurl4-openssl-dev \
    libboost-filesystem-dev && \
    rm -rf /var/lib/apt/lists/*

RUN mkdir -p /app && mkdir -p /.cache
# cache dir for llama.cpp to download models
RUN chmod -R 777 /.cache

WORKDIR /app
RUN git clone https://github.com/ggml-org/llama.cpp.git
# RUN git clone https://github.com/ikawrakow/ik_llama.cpp.git llama.cpp

WORKDIR /app/llama.cpp
RUN cmake -B build \
          -DGGML_LTO=ON \
          -DLLAMA_CURL=ON \
          -DLLAMA_BUILD_SERVER=ON \
          -DLLAMA_BUILD_EXAMPLES=ON \
          -DGGML_ALL_WARNINGS=OFF \
          -DGGML_ALL_WARNINGS_3RD_PARTY=OFF \
          -DCMAKE_BUILD_TYPE=Release
# RUN cmake --build build --config Release --target llama-server -j $(nproc)
RUN cmake --build build --config Release --target llama-server -j 6

WORKDIR /app

EXPOSE 7860

# CMD ["/app/llama.cpp/build/bin/llama-server", "--verbose-prompt"]

# llama-server -hf ggml-org/Qwen3.6-35B-A3B-GGUF

# CMD ["/app/llama.cpp/build/bin/llama-server", "-hf",  "hf.co/ggml-org/Qwen3.6-27B-GGUF:Q4_0",  "--verbose-prompt"]

# ollama run hf.co/ggml-org/Qwen3.6-27B-GGUF:Q4_0
# llama-server -hf ggml-org/Qwen3.6-27B-GGUF:Q4_0
# llama-cli -hf ggml-org/Qwen3.6-27B-GGUF:Q4_0

# llama-cli -hf ggml-org/Qwen3.6-27B-GGUF:Q8_0

# llama-server -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4_K_M molab 14.45tps
# llama-cli -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4_K_M

# unsloth/Qwen3.6-35B-A3B-GGUF:Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf 13.2GB

# CMD ["/app/llama.cpp/build/bin/llama-server", "-hf",  "unsloth/Qwen3.6-35B-A3B-GGUF:Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf"]

# CMD ["/app/llama.cpp/build/bin/llama-server", "-hf", "unsloth/Qwen3.6-35B-A3B-GGUF", "-m", "Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf"]

# llama-server -hf unsloth/Qwen3.6-27B-GGUF:UD-Q4_K_XL
# CMD ["/app/llama.cpp/build/bin/llama-server", "-hf", "unsloth/Qwen3.6-27B-GGUF:UD-Q4_K_XL"]

CMD ["/app/llama.cpp/build/bin/llama-server", "-hf", "unsloth/Qwen3.6-27B-GGUF:UD-IQ3_XXS", "--alias", "qwen3.6-27b-iq3"]

# Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf

# LD_LIBRARY_PATH=/usr/local/cuda-13.0/targets/x86_64-linux/lib:$LD_LIBRARY_PATH ./build/bin/llama-cli -hf unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4_K_M -ngl 99 -p "You are a helpful AI assistant."