Docker Image Size Discrepancy: linux/amd64 6x Larger than linux/arm64

Problem Description

I’m building a Docker image for my Python Flask application on my Mac. I’ve noticed a significant size discrepancy between the linux/amd64 and linux/arm64 targeted images:

• linux/amd64 image size: ~6GB
• linux/arm64 image size: ~1GB

This 6x size difference persists even when building from Docker Cloud or GitHub, ruling out any local environment issues.

Environment

• Host: MacOS (Apple Silicon)
• Python version: 3.11
• Base image: python:3.11-slim

Dockerfile

# Stage 1: Builder
FROM python:3.11-slim AS compile-image

ENV PYTHONDONTWRITEBYTECODE=1
ENV PYTHONUNBUFFERED=1

WORKDIR /app

# Copy requirements first for better cache utilization
COPY ./chatbot/chatbot_requirements.txt .

# Create virtual environment and install dependencies
RUN python -m venv /opt/venv
ENV PATH="/opt/venv/bin:$PATH"
RUN pip install --upgrade pip && 
    pip install --no-cache-dir -r chatbot_requirements.txt && 
    rm -rf /root/.cache/pip chatbot_requirements.txt

# Stage 2: Final
FROM python:3.11-slim AS runtime-image

ENV PYTHONDONTWRITEBYTECODE=1
ENV PYTHONUNBUFFERED=1
ENV PYTHONPATH="/app/:/app/chatbot/"
ENV PATH="/opt/venv/bin:$PATH"

# Create a non-root user
RUN useradd -m appuser

WORKDIR /app

# Copy the virtual environment from the builder stage
COPY --from=compile-image /opt/venv /opt/venv

# Copy application files
COPY --chown=appuser:appuser ./chatbot/ /app/chatbot/
COPY --chown=appuser:appuser .env .

# Switch to non-root user
USER appuser

EXPOSE 8000

CMD ["python", "chatbot/api/chatbot_endpoint.py"]

Requirements

python-dotenv
pydantic
pydantic-settings
pinecone
langchain
langchain-pinecone
langchain-openai
langchain-huggingface
langserve
langgraph
fastapi
uvicorn
sse_starlette

Questions

1. Why is there such a significant size difference between the linux/amd64 and linux/arm64 images?
2. How can I reduce the size of the linux/amd64 image to be closer to that of the linux/arm64 image?

Attempted Solutions

• Tried building on different platforms (local Mac, Docker Cloud, GitHub) with the same results.

Any insights or suggestions would be greatly appreciated!