ContainerFile

#Set Versions.
ARG FEDORA_VERSION=41
ARG OS_TYPE=x86_64
ARG GCC_VERSION=13

#Base image
FROM quay.io/foundata/fedora${FEDORA_VERSION}-itt:latest
USER root

# Use the above args
ARG OS_TYPE
ARG FEDORA_VERSION
ARG GCC_VERSION
ARG PY_VER

ENV HOME=/root
ENV GCC_VERSION=${GCC_VERSION}
ENV CCACHE_DISABLE=1

RUN dnf upgrade --refresh -y && \
    dnf install -y  \
    curl  \
    wget \
    jq \
    vim \
    git \
    gdb \
    unzip \
    which \
    libffi-devel \
    openssl-devel \
    make \
    cmake \
    ninja-build \
    ccache \
    uv \
    zsh \
    findutils \
    python${PY_VER} \
    python${PY_VER}-devel \
    python${PY_VER}-test && \
    dnf clean all

RUN chsh -s $(which zsh)

# Copy the bash script into the image
COPY ./.gitlab/ci/scripts/tools/install_gcc.sh /usr/local/bin/install_gcc_version.sh
RUN bash /usr/local/bin/install_gcc_version.sh

ENV CC=/usr/bin/gcc-13 \
    CXX=/usr/bin/g++-13 \
    CUDAHOSTCXX=/usr/bin/g++-13

RUN mkdir ./cuda-toolkit && \
    cd ./cuda-toolkit && \
    wget https://developer.download.nvidia.com/compute/cuda/12.8.1/local_installers/cuda-repo-fedora41-12-8-local-12.8.1_570.124.06-1.x86_64.rpm && \
    rpm -i cuda-repo-fedora41-12-8-local-12.8.1_570.124.06-1.x86_64.rpm && \
    dnf clean all && \
    dnf -y install cuda-toolkit-12-8 && \
    cd .. && \
    rm -rf ./cuda-toolkit

# CUDA env
ENV CUDA_HOME=/usr/local/cuda
ENV PATH="/usr/local/cuda-12.8/bin:${PATH}"
ENV LD_LIBRARY_PATH="/usr/local/cuda/lib64:${LD_LIBRARY_PATH}"

# NVCC host compiler path
ENV NVCC_CCBIN=/usr/bin/g++
RUN echo 'export NVCC_CCBIN="$(command -v g++)"' > /etc/profile.d/nvcc_ccbin.sh

RUN dnf remove -y xz \
    && cd /tmp \
    && curl -LO https://tukaani.org/xz/xz-5.2.5.tar.gz \
    && tar -xzf xz-5.2.5.tar.gz \
    && cd xz-5.2.5 \
    && ./configure --disable-sandbox \
    && make -j"$(nproc)" \
    && make install \
    && cd /tmp \
    && rm -rf xz-5.2.5* \
    && dnf clean all \
    && rm -rf /var/cache/dnf

#Install cudnn
RUN mkdir ./cudnn
RUN cd ./cudnn \
    && wget https://developer.download.nvidia.com/compute/cudnn/redist/cudnn/linux-x86_64/cudnn-linux-x86_64-9.6.0.74_cuda12-archive.tar.xz \
    && tar xvf cudnn-linux-x86_64-9.6.0.74_cuda12-archive.tar.xz \
    && mv ./cudnn-linux-x86_64-9.6.0.74_cuda12-archive/include/* /usr/local/cuda/include \
    && mv ./cudnn-linux-x86_64-9.6.0.74_cuda12-archive/lib/* /usr/local/cuda/lib64 \
    && cd .. && rm -r ./cudnn

#Update file permissions
RUN chmod a+r /usr/local/cuda/include/cudnn*.h /usr/local/cuda/lib64/libcudnn*

# Create dev user before setting up virtualenv
RUN dnf install -y sudo shadow-utils && dnf clean all
# Fix sudo permissions (setuid bit required for non-root users)
RUN chown root:root /usr/bin/sudo && chmod 4755 /usr/bin/sudo

ARG USERNAME=dev
ARG USER_UID=1011
ARG USER_GID=$USER_UID

RUN groupadd --gid $USER_GID $USERNAME && \
    useradd --uid $USER_UID --gid $USER_GID -m -s /bin/bash $USERNAME && \
    echo "$USERNAME ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers.d/$USERNAME && \
    chmod 0440 /etc/sudoers.d/$USERNAME

# Switch HOME to dev user's home before creating virtualenv
ENV HOME=/home/$USERNAME

# Install Python with uv and setup virtualenv
RUN mkdir -p /pytorch/
COPY ./pytorch/pyproject.toml /pytorch/
RUN cd $HOME && \
    uv python install "${PY_VER}" && \
    uv venv --python "${PY_VER}" --clear $HOME/.venv && \
    bash -c "source $HOME/.venv/bin/activate && uv pip install --upgrade pip" && \
    bash -c "source $HOME/.venv/bin/activate && cd /pytorch && uv pip install --group dev && uv pip install mkl-static mkl-include"


#Install for magma
RUN mkdir -p /pytorch/.ci/docker/common/
COPY ./pytorch/.ci/docker/common/install_magma.sh /pytorch/.ci/docker/common/
RUN bash -c "source $HOME/.venv/bin/activate && cd /pytorch && bash .ci/docker/common/install_magma.sh 12.8"

#Install triton.
RUN mkdir -p /pytorch/scripts
COPY ./pytorch/scripts/install_triton_wheel.sh /pytorch/scripts
RUN mkdir -p /pytorch/.ci/docker/
COPY ./pytorch/.ci/docker/triton_version.txt /pytorch/.ci/docker/triton_xpu_version.txt /pytorch/.ci/docker/
RUN mkdir -p /pytorch/.ci/docker/ci_commit_pins/
COPY ./pytorch/.ci/docker/ci_commit_pins/triton.txt /pytorch/.ci/docker/ci_commit_pins/triton-xpu.txt /pytorch/.ci/docker/ci_commit_pins/
COPY ./pytorch/Makefile /pytorch/
RUN bash -c "source $HOME/.venv/bin/activate && cd /pytorch && make triton"

RUN ldconfig

RUN echo "=== Compiler Verification ===" && \
    gcc --version && \
    g++ --version && \
    nvcc --version && \
    echo "=== C++ Header Directories ===" && \
    ls -la /usr/include/c++/ && \
    echo "=== Environment Variables ===" && \
    echo "CC=$CC" && \
    echo "CXX=$CXX" && \
    echo "NVCC_CCBIN=$NVCC_CCBIN" && \
    echo "CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH"

# Fix ownership of venv created as root
RUN chown -R $USERNAME:$USERNAME $HOME/.venv
# Before switching to dev user
RUN mkdir -p /home/dev && \
    chown -R $USERNAME:$USERNAME /home/dev
USER $USERNAME