docker/Dockerfile

# DreamZero 训练 Docker 镜像
#
# 构建:
#   docker build -t dreamzero:latest -f docker/Dockerfile .
#
# 运行 (单节点 8 GPU):
#   docker run --gpus all --shm-size=64g \
#     -v /path/to/checkpoints:/workspace/dreamzero/checkpoints \
#     -v /path/to/data:/workspace/dreamzero/data \
#     -v /path/to/output:/workspace/dreamzero/output \
#     dreamzero:latest \
#     data=dreamzero/libero \
#     train_architecture=full \
#     max_steps=200000 \
#     output_dir=./output/libero_full

FROM nvidia/cuda:12.4.1-devel-ubuntu22.04

LABEL description="DreamZero: Video Diffusion Model for Scalable Robot Learning"
LABEL maintainer="DreamZero Team"

# ============ 系统依赖 ============
RUN apt-get update && apt-get install -y --no-install-recommends \
    python3.11 \
    python3.11-dev \
    python3.11-distutils \
    python3-pip \
    git \
    ffmpeg \
    libsm6 \
    libxext6 \
    libglib2.0-0 \
    libxrender-dev \
    libgl1-mesa-glx \
    wget \
    curl \
    && rm -rf /var/lib/apt/lists/*

# 设置 python3.11 为默认
RUN update-alternatives --install /usr/bin/python python /usr/bin/python3.11 1
RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.11 1

# ============ 安装 PyTorch ============
RUN pip install --no-cache-dir \
    torch==2.4.0 \
    torchvision==0.19.0 \
    --index-url https://download.pytorch.org/whl/cu124

# ============ 复制代码 ============
WORKDIR /workspace/dreamzero
COPY groot/ ./groot/
COPY scripts/ ./scripts/
COPY pyproject.toml requirements.txt ./

# ============ 安装依赖 ============
RUN pip install --no-cache-dir -e ".[train]"

# ============ NCCL 配置（多节点优化） ============
ENV NCCL_DEBUG=WARN
ENV NCCL_IB_DISABLE=0
ENV NCCL_SOCKET_IFNAME=^docker0,lo
ENV NCCL_IB_TIMEOUT=22
ENV NCCL_IB_RETRY_CNT=4

# ============ 中国用户镜像 ============
ENV HF_ENDPOINT=https://hf-mirror.com

# ============ 默认入口 ============
ENTRYPOINT ["torchrun", "--standalone", "--nproc_per_node", "8", \
    "/workspace/dreamzero/groot/vla/experiment/experiment.py"]
CMD ["--help"]