Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Using Adapter Transformers at Hugging Face

adapter-transformers is a library that extends 🤗 transformers by allowing to integrate, train and use Adapters and other efficient fine-tuning methods. The library is fully compatible with 🤗 transformers. Adapters are small learnt layers inserted within each layer of a pre-trained model. You can learn more about this in the original paper.

Exploring adapter-transformers in the Hub

You can find over a hundred adapter-transformer models by filtering at the left of the models page. Some adapter models can be found in the Adapter Hub repository. Models from both sources are then aggregated in the AdapterHub.

Using existing models

For a full guide on loading pre-trained adapters, we recommend checking out the official guide.

As a brief summary, once you load a model with the usual *Model classes from 🤗transformers, you can use the load_adapter method to load and activate the Adapter (remember adapter-transformers extends 🤗transformers.).

from transformers import AutoModelWithHeads

model = AutoModelWithHeads.from_pretrained("bert-base-uncased")
adapter_name = model.load_adapter("AdapterHub/bert-base-uncased-pf-imdb", source="hf")
model.active_adapters = adapter_name

You can also use list_adapters to find all Adapter Models programmatically

from transformers import list_adapters

# source can be "ah" (AdapterHub), "hf" ( or None (for both, default)
adapter_infos = list_adapters(source="hf", model_name="bert-base-uncased")

If you want to see how to load a specific model, you can click Use in Adapter Transformers and you will be given a working snippet that you can load it!

Sharing your models

For a full guide on sharing models with adapter-transformers, we recommend checking out the official guide.

You can share your Adapter by using the push_adapter_to_hub method from a model that already contains an adapter.


This command creates a repository with an automatically generated model card and all necessary metadata.

Additional resources