move from /root to /opt/nvfp4

Dockerfile

@@ -1,11 +1,14 @@
 FROM nvidia/cuda:13.2.1-cudnn-devel-ubuntu24.04
 ENV DEBIAN_FRONTEND=noninteractive
+
+# "Unminify"
 RUN rm -f /etc/dpkg/dpkg.cfg.d/excludes && \
     rm -f /etc/apt/apt.conf.d/docker-clean && \
     rm -f /etc/apt/apt.conf.d/01-nodoc && \
     rm -f /usr/bin/man && \
     dpkg-divert --quiet --remove /usr/bin/man
 
+# Install packages
 RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
     --mount=type=cache,target=/var/lib/apt,sharing=locked \
     apt update && apt-get dist-upgrade --autoremove -y && apt-get install -y \
@@ -21,15 +24,23 @@ RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
     git \
     cmake
 
-RUN python3 -m venv /root/venv
+# Set up nvfp4 python virtual environment
-ENV PATH="/root/venv/bin:${PATH}"
+WORKDIR /opt/nvfp4
+RUN python3 -m venv /opt/nvfp4/py-venv
+ENV PATH="/opt/nvfp4/py-venv/bin:${PATH}"
 RUN --mount=type=cache,target=/root/.cache/pip,sharing=locked pip3 install torch torchvision torchaudio --upgrade --index-url \
     https://download.pytorch.org/whl/cu132/
 
-WORKDIR /root
+
-RUN mkdir llama.cpp
+# Fetch convert_to_quant, and install it into our python venv.
-COPY --exclude=build/ llama.cpp ./llama.cpp/
+WORKDIR /opt/nvfp4
-WORKDIR /root/llama.cpp
+RUN git clone --depth=1 https://github.com/silveroxides/convert_to_quant && cd convert_to_quant && rm -rf .git && pip3 install -r requirements.txt && pip3 install .
+
+# Clone and compile llama.cpp (save some space by removing unneeded .git data)
+RUN git clone --depth=1 https://github.com/ggml-org/llama.cpp && rm -rf llama.cpp/.git
+WORKDIR /opt/nvfp4/llama.cpp
+
+# Build llama.cpp, the ldconfig is important
 RUN echo "/usr/local/cuda-13.2/compat" > /etc/ld.so.conf.d/988-cuda-compat.conf && ldconfig
 RUN cmake -B build -S . \
      -DCMAKE_BUILD_TYPE=Release \
@@ -41,22 +52,23 @@ RUN cmake -B build -S . \
     && cmake --build build --target install \
     && rm -rf build
 
-WORKDIR /root
+# Copy and install zefie's nvfp4 quant scripts
-RUN git clone https://github.com/silveroxides/convert_to_quant && cd convert_to_quant && pip3 install -r requirements.txt && pip3 install .
-
 RUN mkdir nvfp4_gguf_quant
-WORKDIR /root/nvfp4_gguf_quant
+WORKDIR /opt/nvfp4/nvfp4_gguf_quant
 COPY --exclude=__pycache__ nvfp4gguf ./nvfp4gguf
 COPY --chmod=0644 gguf_to_nvfp4_gguf.py hf_to_nvfp4_gguf.py hf_to_nvfp4_safetensors.py requirements.txt safetensors_to_nvfp4_gguf.py hf_info.py gguf_info.py .
 
+# Install requirements for zefie's nvfp4 quant scripts
 RUN --mount=type=cache,target=/root/.cache/pip,sharing=locked pip3 install -r requirements.txt
 COPY --chmod=0644 docker/README.txt /workspace/README.txt
 COPY --chmod=0755 docker/hf_info docker/gguf_info docker/hf_to_nvfp4_gguf docker/hf_to_nvfp4_safetensors docker/gguf_to_nvfp4_gguf docker/safetensors_to_nvfp4_gguf /usr/local/bin/
 
+# Install manpages for zefie's nvfp4 quant scripts, ctq, and heretic
 RUN mkdir -p /usr/local/share/man/man1
 COPY --chmod=0644 man/man1/* /usr/local/share/man/man1/
-
 RUN --mount=type=cache,target=/root/.cache/pip,sharing=locked pip3 install heretic-llm
 
+# Finalize and settle at /workspace
+ENV DEBIAN_FRONTEND=
 WORKDIR /workspace
-
+ENTRYPOINT []

docker/gguf_info

@@ -1,2 +1,2 @@
 #!/bin/bash
-python /root/nvfp4_gguf_quant/gguf_info.py "${@}"
+python /opt/nvfp4/nvfp4_gguf_quant/gguf_info.py "${@}"

docker/gguf_to_nvfp4_gguf

@@ -1,2 +1,2 @@
 #!/bin/bash
-python /root/nvfp4_gguf_quant/gguf_to_nvfp4_gguf.py --llama-cpp /root/llama.cpp "${@}"
+python /opt/nvfp4/nvfp4_gguf_quant/gguf_to_nvfp4_gguf.py --llama-cpp /opt/nvfp4/llama.cpp "${@}"

docker/hf_info

@@ -1,2 +1,2 @@
 #!/bin/bash
-python /root/nvfp4_gguf_quant/hf_info.py "${@}"
+python /opt/nvfp4/nvfp4_gguf_quant/hf_info.py "${@}"

docker/hf_to_nvfp4_gguf

@@ -1,2 +1,2 @@
 #!/bin/bash
-python /root/nvfp4_gguf_quant/hf_to_nvfp4_gguf.py --llama-cpp /root/llama.cpp "${@}"
+python /opt/nvfp4/nvfp4_gguf_quant/hf_to_nvfp4_gguf.py --llama-cpp /opt/nvfp4/llama.cpp "${@}"

docker/hf_to_nvfp4_safetensors

@@ -1,2 +1,2 @@
 #!/bin/bash
-python /root/nvfp4_gguf_quant/hf_to_nvfp4_safetensors.py --convert-to-quant /root/convert_to_quant "${@}"
+python /opt/nvfp4/nvfp4_gguf_quant/hf_to_nvfp4_safetensors.py --convert-to-quant /opt/nvfp4/convert_to_quant "${@}"

docker/safetensors_to_nvfp4_gguf

@@ -1,2 +1,2 @@
 #!/bin/bash
-python /root/nvfp4_gguf_quant/safetensors_to_nvfp4_gguf.py --llama-cpp /root/llama.cpp "${@}"
+python /opt/nvfp4/nvfp4_gguf_quant/safetensors_to_nvfp4_gguf.py --llama-cpp /opt/nvfp4/llama.cpp "${@}"