Hub

TF-Keras (legacy)

tf-keras is the name given to Keras 2.x version. It is now hosted as a separate GitHub repo here. Though it’s a legacy framework, there are still 4.5k+ models hosted on the Hub. These models can be loaded using the huggingface_hub library. You must have either tf-keras or keras<3.x installed on your machine.

If you are interested in Keras 3.x support, check out this guide.

Once installed, you just need to use the from_pretrained_keras method to load a model from the Hub. Read more about from_pretrained_keras here.

from huggingface_hub import from_pretrained_keras

model = from_pretrained_keras("keras-io/mobile-vit-xxs")
prediction = model.predict(image)
prediction = tf.squeeze(tf.round(prediction))
print(f'The image is a {classes[(np.argmax(prediction))]}!')

# The image is a sunflower!

You can also host your tf-keras model on the Hub. However, keep in mind that tf-keras is a legacy framework. To reach a maximum number of users, we recommend to create your model using Keras 3.x and share it natively as described above. For more details about uploading tf-keras models, check out push_to_hub_keras documentation.

from huggingface_hub import push_to_hub_keras

push_to_hub_keras(model,
    "your-username/your-model-name",
    "your-tensorboard-log-directory",
    tags = ["object-detection", "some_other_tag"],
    **model_save_kwargs,
)

Additional resources

• GitHub repo

• Blog post Putting Keras on 🤗 Hub for Collaborative Training and Reproducibility (April 2022)