Implementation:Tensorflow Serving Docker Build Serving Image

Knowledge Sources	TensorFlow Serving TF Serving Kubernetes
Domains	Containerization, Deployment
Last Updated	2026-02-13 17:00 GMT

Overview

Concrete tool for building a Docker image with a baked-in model using docker run, docker cp, and docker commit commands.

Description

The TensorFlow Serving Kubernetes tutorial uses a three-step Docker build:

1. docker run -d: Start a detached container from tensorflow/serving base image
2. docker cp: Copy the SavedModel directory into the container at /models/{model_name}
3. docker commit --change "ENV MODEL_NAME {name}": Commit as a new image

The base image entrypoint script runs: tensorflow_model_server --port=8500 --rest_api_port=8501 --model_name=${MODEL_NAME} --model_base_path=${MODEL_BASE_PATH}/${MODEL_NAME}

Usage

Run these Docker commands after preparing the SavedModel locally. The resulting image is ready for registry push and Kubernetes deployment.

Code Reference

Source Location

• Repository: tensorflow/serving
• File: tensorflow_serving/g3doc/serving_kubernetes.md (L52-84)

Signature

# Step 1: Start temporary container
docker run -d --name serving_base tensorflow/serving

# Step 2: Copy model into container
docker cp /tmp/resnet serving_base:/models/resnet

# Step 3: Commit as new image
docker commit --change "ENV MODEL_NAME resnet" serving_base $USER/resnet_serving

# Cleanup
docker kill serving_base && docker rm serving_base

Import

# Requires: docker CLI
# Base image: tensorflow/serving (from Docker Hub)

I/O Contract

Inputs

Name	Type	Required	Description
Local SavedModel path	string	Yes	Path to the model directory (e.g., /tmp/resnet)
Model name	string	Yes	Name for the model (sets MODEL_NAME env var)
Base image	string	No	Default: tensorflow/serving

Outputs

Name	Type	Description
Docker image	image	$USER/resnet_serving with model baked in at /models/{name}

Usage Examples

Build ResNet Serving Image

# Build image with ResNet model
docker run -d --name serving_base tensorflow/serving
docker cp /tmp/resnet serving_base:/models/resnet
docker commit --change "ENV MODEL_NAME resnet" serving_base $USER/resnet_serving
docker kill serving_base && docker rm serving_base

# Test locally
docker run -p 8500:8500 -p 8501:8501 $USER/resnet_serving

Related Pages

Implements Principle

• Principle:Tensorflow_Serving_Docker_Image_Building

Requires Environment

• Environment:Tensorflow_Serving_Docker_Runtime_Environment