How to Create Dockerfile for ML Model: Simple Guide

To create a Dockerfile for an ML model, start with a base image like python, copy your model and code files, install dependencies using pip, and define the command to run your model. This packages your model and environment so it runs consistently anywhere with docker build and docker run.

Syntax

A Dockerfile is a text file with instructions to build a Docker image. Key parts include:

• FROM: base image to start from (e.g., python)
• COPY: copy your code and model files into the image
• RUN: run commands like installing packages
• WORKDIR: set working directory inside the container
• CMD: command to run your ML model when container starts

dockerfile

FROM python:3.9-slim
WORKDIR /app
COPY requirements.txt ./
RUN pip install --no-cache-dir -r requirements.txt
COPY . ./
CMD ["python", "app.py"]

Example

This example Dockerfile packages a simple ML model app. It uses Python 3.9, installs dependencies from requirements.txt, copies all files, and runs app.py which loads and serves the model.

dockerfile

FROM python:3.9-slim
WORKDIR /app
COPY requirements.txt ./
RUN pip install --no-cache-dir -r requirements.txt
COPY . ./
CMD ["python", "app.py"]

Output

Successfully built image and container runs your ML model app

Common Pitfalls

Common mistakes when creating Dockerfiles for ML models include:

• Not specifying a lightweight base image, causing large images
• Forgetting to copy all necessary files (model weights, code)
• Not installing dependencies properly, leading to runtime errors
• Using RUN python app.py instead of CMD, which runs at build time instead of container start

dockerfile

### Wrong way (runs app during build, not container start)
FROM python:3.9-slim
WORKDIR /app
COPY requirements.txt ./
RUN pip install -r requirements.txt
COPY . ./
RUN python app.py

### Right way
FROM python:3.9-slim
WORKDIR /app
COPY requirements.txt ./
RUN pip install -r requirements.txt
COPY . ./
CMD ["python", "app.py"]

Quick Reference

Tips for Dockerfile creation for ML models:

• Use slim or minimal base images to reduce size
• Pin dependency versions in requirements.txt for reproducibility
• Copy only needed files to keep image clean
• Use CMD to specify the runtime command
• Test your container locally before deployment

Key Takeaways

Start your Dockerfile with a lightweight Python base image like python:3.9-slim.

Copy your ML model files and install dependencies using pip inside the Dockerfile.

Use CMD to specify the command that runs your ML model when the container starts.

Avoid running your app during image build; run it when the container runs.

Keep your Docker image small and reproducible by pinning dependency versions.