Ctrl+K

Getting started with teras

Getting started with teras#

teras strives to make tabular deep learning accessible. It goes without saying, but since teras is based on Keras so all layers and models available in teras are just wrappers around Keras layers and models and hence provide seamless integration.

Note

teras v0.3 is now fully based on Keras 3, making everything available backend agnostic. It supports TensorFlow, JAX and PyTorch backends.

Warning

To use teras v0.3 you must have Keras 3 installed! It won’t work with Keras 2.x

Installing teras#

You can install teras using pip as follows,

>>> pip install teras

Configuring backend#

You can export the environment variable KERAS_BACKEND or you can edit your local config file at ~/.keras/keras.json to configure your backend. Available backend options are: “jax”, “tensorflow”, “torch”. Example:

>>> export KERAS_BACKEND="jax"

In Colab, you can do,

import os
os.environ["KERAS_BACKEND"] = "jax"
import keras

For more Keras related configuration, please refer to Getting started with Keras.

Contents