Implementation:Tensorflow Tfjs Tensorflowjs Pip Install

Knowledge Sources	TensorFlow.js tensorflowjs PyPI TensorFlow.js Converter Guide
Domains	Tooling, Deployment
Principle	Principle:Tensorflow_Tfjs_Converter_Installation
Type	External Tool Doc
Last Updated	2026-02-10 00:00 GMT

Environment:Tensorflow_Tfjs_Python_Converter

Overview

This implementation documents the concrete installation procedure for the tensorflowjs Python pip package, which provides the tensorflowjs_converter CLI tool and the tensorflowjs_wizard interactive tool. These tools are used to convert Python TensorFlow models into the TensorFlow.js format for browser and Node.js deployment.

Installation Command

Basic Installation

pip install tensorflowjs

Version-Pinned Installation

# Install a specific version matching your target TF.js runtime
pip install tensorflowjs==4.17.0

# Install the latest version
pip install --upgrade tensorflowjs

Virtual Environment Installation (Recommended)

# Create and activate a virtual environment
python -m venv tfjs_env
source tfjs_env/bin/activate  # Linux/macOS
# tfjs_env\Scripts\activate   # Windows

# Install tensorflowjs within the virtual environment
pip install tensorflowjs

Conda Environment Installation

# Create a conda environment
conda create -n tfjs python=3.10
conda activate tfjs

# Install via pip within the conda environment
pip install tensorflowjs

Inputs and Outputs

Inputs

A Python environment (Python 3.9 or later recommended) with pip available
Network access to the Python Package Index (PyPI) at pypi.org
Optionally, a specific version number for reproducible builds

Outputs

After successful installation, the following tools are available:

Tool	Location	Description
tensorflowjs_converter	In PATH (e.g., ~/.local/bin/ or venv/bin/)	CLI tool for batch model conversion
tensorflowjs_wizard	In PATH	Interactive CLI wizard that guides through conversion options
tensorflowjs Python module	Site-packages	Python API for programmatic conversion

Verification

After installation, verify the tools are correctly installed and accessible:

# Verify the converter CLI is available
tensorflowjs_converter --version

# Verify the wizard is available
tensorflowjs_wizard --help

# Verify the Python module is importable
python -c "import tensorflowjs; print(tensorflowjs.__version__)"

# List the full set of converter options
tensorflowjs_converter --help

Expected output from --help includes usage information showing all supported --input_format and --output_format values, quantization flags, and other conversion options.

Dependencies

The tensorflowjs package installs the following key dependencies (versions depend on the tensorflowjs version):

Dependency	Purpose	Notes
tensorflow	Core TF runtime for reading model formats	Major version must be compatible with tensorflowjs version
tensorflow-hub	Support for TF Hub module conversion	Required for --input_format=tf_hub
numpy	Numerical operations on weight data	Used for weight data manipulation during conversion
six	Python 2/3 compatibility utilities	Transitive dependency
h5py	HDF5 file reading/writing	Required for --input_format=keras (HDF5 models)
packaging	Version parsing and comparison	For version compatibility checks

Version Compatibility Matrix

tensorflowjs Version	TensorFlow Version	TF.js Runtime Version	Python Version
4.17.x	2.15.x	4.17.x	3.9 - 3.11
4.10.x - 4.16.x	2.13.x - 2.15.x	4.10.x - 4.16.x	3.9 - 3.11
4.0.x - 4.9.x	2.11.x - 2.13.x	4.0.x - 4.9.x	3.8 - 3.11
3.x	2.8.x - 2.10.x	3.x	3.7 - 3.10

Note: Always check the PyPI page for the latest compatibility information, as these ranges evolve with each release.

The tensorflowjs_wizard

The tensorflowjs_wizard provides an interactive alternative to the CLI converter:

tensorflowjs_wizard

The wizard prompts for:

Input model format — Select from the list of supported formats
Input model path — Path to the source model
Output format — Select the target TF.js format
Output directory — Where to write the converted model
Quantization — Whether and how to quantize weights
Weight sharding — Shard size configuration

This is useful for one-off conversions or for users unfamiliar with the CLI flags.

Programmatic Usage

The tensorflowjs Python module can also be used programmatically for conversion within Python scripts or CI/CD pipelines:

import tensorflowjs as tfjs

# Convert a Keras model object directly (no file I/O for input)
import tensorflow as tf
model = tf.keras.applications.MobileNetV2(weights='imagenet')
tfjs.converters.save_keras_model(model, '/tmp/tfjs_mobilenet')

# Convert from SavedModel path
tfjs.converters.convert_tf_saved_model(
    '/tmp/my_saved_model',
    '/tmp/tfjs_output',
    signature_def='serving_default',
    saved_model_tags='serve'
)

Common Issues

Issue	Cause	Resolution
Command not found: tensorflowjs_converter	pip installed to a path not in PATH	Use pip show tensorflowjs to find the install location; add the scripts directory to PATH, or use python -m tensorflowjs.converters.converter_wrapper
TensorFlow version mismatch	tensorflowjs requires a specific TF version	Install a compatible TensorFlow version, or upgrade/downgrade tensorflowjs
Permission denied during install	System-level pip without sudo	Use pip install --user tensorflowjs or install in a virtual environment
ImportError: No module named tensorflowjs	Python environment mismatch	Ensure you are using the same Python environment where tensorflowjs was installed
GPU-related errors	TensorFlow GPU build conflicts	Use pip install tensorflow-cpu if GPU support is not needed for conversion

CI/CD Integration

For automated pipelines, pin the exact version and include verification:

# requirements-converter.txt
tensorflowjs==4.17.0

# CI step
pip install -r requirements-converter.txt
tensorflowjs_converter --version  # Verify installation
tensorflowjs_converter \
    --input_format=tf_saved_model \
    --output_format=tfjs_graph_model \
    /path/to/saved_model \
    /path/to/output

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment