tf-keras

What it is and what it does

TF-Keras is the TensorFlow-native implementation of Keras, descended from the legacy `tf.keras` codebase. It provides a high-level, user-friendly API for defining, training, and evaluating deep neural networks using TensorFlow as the computational backend. The package includes layers, optimizers, loss functions, and utilities for building models ranging from simple sequential architectures to complex custom networks.

Unlike the newer Keras 3 (multi-backend implementation supporting JAX, PyTorch, and TensorFlow), TF-Keras is specifically optimized for TensorFlow workflows. It is production-stable and actively maintained, making it suitable for projects where TensorFlow is the primary or exclusive deep learning framework.

Use it for:

• Build and train sequential or functional neural network models for image classification, NLP, or regression tasks.
• Rapidly prototype deep learning models using high-level Keras layers and pre-built components.
• Fine-tune pre-trained models or transfer learning workflows within a TensorFlow ecosystem.
• Define custom layers and loss functions for specialized deep learning research or production systems.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

TF-Keras is the pure-TensorFlow implementation of Keras, providing a high-level API for building and training deep learning models with TensorFlow as the backend.

Yes, if you are committed to TensorFlow-based deep learning. TF-Keras is production-stable, actively maintained, has low install friction, and carries a permissive Apache 2.0 license. No known vulnerabilities. Install it as the standard Keras API for TensorFlow projects; consider Keras 3 only if you need multi-backend flexibility.

Install

tf-keras on PyPI

pip install tf-keras

uv add tf-keras

poetry add tf-keras

Installing tf-keras

Before you install

Low install friction with a pure Python wheel. Actively maintained with a recent commit on 2026-08-14 and production-stable status. Requires Python 3.10 or later.

License in practice

Licensed under Apache 2.0 (permissive), allowing commercial use, modification, and distribution with minimal restrictions.

Quickstart

pip install tf-keras

import tensorflow as tf
from tensorflow import keras

model = keras.Sequential([keras.layers.Dense(10, activation='relu')])
model.compile(optimizer='adam', loss='mse')

Requires Python 3.10 or later and TensorFlow as a runtime dependency.

Verify before relying

• Whether TF-Keras 2.21.0 is compatible with all current TensorFlow versions or has specific version constraints.
• Performance characteristics compared to the multi-backend Keras 3 implementation.

Package facts

Evidence: tf_keras-2.21.0-py3-none-any.whl

Keywords: keras, tensorflow, machine learning, deep learning

Tags

More Software Development packages