H-JEPA/Dockerfile.inference at main · jonwiggins/H-JEPA · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
# Inference Dockerfile for H-JEPA
# Optimized for production serving with minimal image size

FROM nvidia/cuda:11.8.0-cudnn8-runtime-ubuntu22.04 AS base

# Prevent interactive prompts
ENV DEBIAN_FRONTEND=noninteractive
ENV PYTHONUNBUFFERED=1

WORKDIR /app

# Install system dependencies (minimal for inference)
RUN apt-get update && apt-get install -y \
    python3.11 \
    python3-pip \
    libglib2.0-0 \
    libsm6 \
    libxext6 \
    libxrender-dev \
    libgomp1 \
    && rm -rf /var/lib/apt/lists/*

# Upgrade pip
RUN python3.11 -m pip install --no-cache-dir --upgrade pip setuptools wheel

# Install PyTorch (CPU or CUDA runtime)
ARG TORCH_DEVICE=cpu
RUN if [ "$TORCH_DEVICE" = "cuda" ]; then \
    pip install --no-cache-dir torch==2.1.0 torchvision==0.16.0 --index-url https://download.pytorch.org/whl/cu118; \
    else \
    pip install --no-cache-dir torch==2.1.0 torchvision==0.16.0 --index-url https://download.pytorch.org/whl/cpu; \
    fi

# Copy requirements and install minimal dependencies for inference
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Install serving dependencies
RUN pip install --no-cache-dir \
    fastapi==0.104.1 \
    uvicorn[standard]==0.24.0 \
    pydantic==2.5.0 \
    python-multipart==0.0.6 \
    aiofiles==23.2.1 \
    prometheus-client==0.19.0

# Copy only necessary source files
COPY src/ ./src/
COPY pyproject.toml .
COPY README.md .

# Install h-jepa package
RUN pip install --no-cache-dir -e .

# Create model directory
RUN mkdir -p /app/models

# Copy model server
COPY src/serving/ ./src/serving/

# Set non-root user for security
RUN useradd -m -u 1000 appuser && \
    chown -R appuser:appuser /app
USER appuser

# Expose API port
EXPOSE 8000

# Health check
HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
    CMD python3.11 -c "import requests; requests.get('http://localhost:8000/health')"

# Default command: run model server
CMD ["python3.11", "-m", "uvicorn", "src.serving.model_server:app", "--host", "0.0.0.0", "--port", "8000"]